Embed Size (px)
Citation preview
Persistent Poverty andLifetime Inequality: The Evidence
Proceedings from a workshop held at H M Treasury,chaired by Professor John Hills,
Director of the ESRC Research Centre for Analysis of Social Exclusion, LSE
17th and 18th November 1998
CENTRE FOR ANALYSIS OF SOCIAL EXCLUSION
An ESRC Research Centre
CASEreport 5March 1999
ISSN 1465-300 1
ASE
Occasional paper no. 10
HM Treasury
Centre for Analysis of Social Exclusion
The ESRC Research Centre for Analysis of Social Exclusion (CASE) was established in October 1997 with funding from theEconomic and Social Research Council. It is located within the Suntory and Toyota International Centres for Economics andRelated Disciplines (STICERD) at London School of Economics and Political Science, and benefits from support fromSTICERD. It is directed by Howard Glennerster, John Hills, Kathleen Kiernan, Julian Le Grand and Anne Power.
Our Discussion Papers series is available free of charge. We also produce summaries of our research in CASEbriefs. Tosubscribe to the series, or for further information on the work of the Centre and our seminar series, please contact the CentreAdministrator, Jane Dickson, on:
Telephone: +44 (0)171 955 6679Fax: +44 (0)171 242 2357Email: [email protected]: sticerd.lse.ac.uk/case.htm
HMT Public Enquiry Unit:
Telephone: +44 (0)171 270 4558Fax: +44 (0)171 270 5244
i
Background
This report summarises presentations and discussion at a workshop on ‘Persistent Poverty and Lifetime Inequality’ organisedby HM Treasury and chaired by John Hills, Director of the ESRC Resarch Centre for Analysis of Social Exclusion at theLondon School of Economics. It took place on 17 and 18 November 1998. A list of presenters is included overleaf. Theorganisers are very grateful to all participants for their contributions to the debate summarised here.
The Treasury decided to hold this workshop to encourage debate and extend understanding of the causes of persistent povertyand inequality of opportunity, drawing on the large amount of new research using panel datasets*. These new datasets make itpossible to move from a static analysis of poverty and inequality to a dynamic focus. Looking at the dynamics of poverty andinequality of opportunity enables us to pinpoint the processes and events which lead people to be at greater risk of low incomeand poorer life chances. These data provide a much firmer underpinning for policies which aim to tackle these problems atsource.
* Several of the papers included in this volume report results drawn from datasets including the British Household Panel Survey, the National Child DevelopmentStudy, the Labour Force Survey and others supplied by the Data Archive at Essex University, to whom the presenters and organisers are most grateful.
ii
Contents Page
Session 1: Poverty and Inequality From a Dynamic Perspective
John Hills: ‘Introduction: What do we mean by reducing lifetime inequality and increasing mobility?’ 1
Stephen Jenkins: ‘Income dynamics’ 5
Robert Walker: ‘Life-cycle trajectories’ 11
Steve Machin: ‘Inter-generational inequality’ 19
Howard Oxley: ‘Income dynamics: inter-generational evidence’ 24
Discussion 30
Session 2: Area and Multiple Deprivation
Sam Mason: ‘The pattern of area deprivation’ 31
Anne Power: ‘The relationship between inequality and area deprivation’ 38
Michael Noble: ‘Longitudinal analysis of area deprivation’ 46
Discussion 50
Session 3: Retirement and the Elderly
Nigel Campbell: ‘Older workers and the labour market’ 52
Richard Disney: ‘Prioritising older workers’ 62
Carl Emmerson: ‘Retirement Incomes’ 67
Discussion 71
Session 4: Work and Poverty
Mark Stewart: ‘Low pay, no pay dynamics’ 73
Richard Dickens: ‘Wage mobility’ 79
Rebecca Endean: ‘Work, low pay and poverty, evidence from the BHPS and LLMDB’ 83
Paul Gregg: ‘Scarring effects of unemployment’ 91
Discussion 99
Session 5: Childhood Poverty and Family Structure
John Bynner: ‘NCDS evidence on early years’ 101
Jane Waldfogel: ‘Childcare and outcomes’ 105
Kathleen Kiernan:‘Divorce / family breakdown’ 113
John Hobcraft*: ‘Intergenerational and Life-Course Transmission of Social Exclusion’ 117
Discussion 122
Session 6: Education and Poverty
Ralph Tabberer:‘Childhood poverty and school attainment, causal effect and impact on lifetime inequality’ 123
David Soskice - Childhood poverty and post-compulsory participation and attainment,causal effect and impact on lifetime inequality 129
Discussion 134
*Presented by Kathleen Kiernan in John Hobcraft’s absence.
iii
List of presentersProfessor John HillsESRC Research Centre for Analysis of Social ExclusionLondon School of Economics
Professor Stephen JenkinsESRC Research Centre on Microsocial ChangesInstitute of Social and Economic ResearchUniversity of Essex
Professor Robert WalkerCentre for Research in Social PolicyDepartment of Social SciencesLoughborough University
Steve MachinUniversity College of London and CEP
Howard Oxley,OECD
Sam Mason Department of the Environment, Transport and Regions
Professor Anne PowerDepartment of Social Policy and Centre for Analysis of Social ExclusionLondon School of Economics
Professor Michael NobleDepartment of Applied Social StudiesOxford University
Nigel CampbellCabinet Office
Professor Richard DisneySchool of EconomicsUniversity of Nottingham
Dr Carl EmmersonInstitute of Fiscal Studies
Professor Mark StewartEconomics DepartmentUniversity of Warwick
Dr Richard DickensCentre for Economic PerformanceLondon School of Economics
Rebecca EndeanDepartment of Social Security
Dr Paul GreggLondon School of Economics/HMT
Professor John BynnerCentre of Longitudinal Studies Institute of Education
Professor Jane WaldfogelColumbia University and Centre for Analysis of Social ExclusionLondon School of Economics
Dr Kathleen KiernanDepartment of Social Policy and Centre for Analysis of Social ExclusionLondon School of Economics
John HobcraftLondon School of Economics
Ralph TabbererDepartment for Education and Employment
David SoskiceNo.10 Downing Street
1
Session 1: Poverty and Inequality from a Dynamic Perspective
Introduction: ‘What do we mean by reducing lifetime inequality and increasing mobility?’
John Hills, Centre for Analysis of Social Exclusion
The title suggested by the conference organisers was an intriguing one, as answers to it affect what sort of issues we should beconcentrating on in looking at the other papers in this collection. Perhaps more specifically, it could be rephrased as: ‘Why –and in what ways – might policy focus on reducing lifetime inequality and increasing mobility?’
Why might policy focus on reducing lifetimeinequality as opposed to just straight forward inequality? If the focus is purelyon lifetime inequality, it implies that we prefer the distribution given by the trajectories shown in (a) rather than those in (b) inFigure 1. In other words, cross-sectional inequality is all right as long as it averages out in the end. ‘We should take the roughwith the smooth’, and not fuss too much about short-term fluctuations. There is clearly some truth in this. But normallyeconomists worry about risk aversion– people often are prepared to pay an insurance premium to reduce the chance ofwobbles like those in trajectory (a).
Figure 1 Figure 2
Such a focus only on lifetime totals also implies that we do not mind how people arrange their total resources over theirlifetime. It is up to them if they want to be poor students for several years and earn more later, or if they want to take a careerbreak financed from savings or borrowing. This seems fair enough. But we doappear to mind in other circumstances. Weare, for instance, prepared to go to a lot of lengths to encourage – or even force – people’s living standards to look liketrajectory (c) rather than trajectory (d) in Figure 2, even if left to themselves they choose (d). We are not prepared to see thecross-sectional inequality implied by (d) evenif it has no lifetimeeffect.
So a preference for lifetime inequality over cross-sectional inequality does not seem to be unambiguous. However, anotherreason for a focus on lifetimeinequality rather than cross-sectional inequality is that it implies a concern with the longer-termeffects of policy. We are prepared to go for trajectory (e) rather than trajectory (f) in Figure 3.
This kind of preference lies behind the ‘hand-up’ not ‘hand-out’ philosophy, and to welfare-to-work being preferred to allround benefit increases by the present Government, for instance. In other words, this suggests a preference for policies withpersistenteffects on later outcomes. These can perhaps be divided into two:
• Policies which change individual characteristicsin a way which lasts, such as education or training.
• But also, if there is “state dependence” in the system, changing the right initial state has long term effect. Do jobs lead tojobs? Which parts of childhood disadvantage have knock-on effects into adulthood? At the other end of working lives, canyou break the cycle of low skills leading to early retirement, leading in turn to low income in retirement?
There is another problem with ‘reducing lifetime inequality’ as a policy target: you cannot measure it. A few years ago, JaneFalkingham and I with other LSE colleagues carried out an exercise looking at this question, building a lifetime simulation
(a) (b) (c) (d)
(e) (f)
Figure 3
2
model, LIFEMOD1. In our hypothetical model of lives ‘if they were all lived continuously in 1985’ we found a Ginicoefficient of 0.20 for lifetime equivalised net income compared to 0.30 for cross-sectional equivalised net income in the samemodel. But these are only results from an analytical model. The world will not ever look like LIFEMOD, because the worldwon’t look like 1985 forever (perhaps luckily). And it is hard to see results from a model being very convincing to theelectorate: ‘You may think the country has become more unequal, but a computer model at LSE says that on a hypotheticallifetime basis, it has not…’ Unfortunately, you have to wait a very long time to get actual lifetime inequality – even if theESRC carries on funding the British Household Panel Survey for another 60 years. Even then, the observed lifetime totalswould be the product of a whole series of different policy mixes and economic environments, which would make it hard toassess policy at one point even in the distant past.
So in operational terms, even if lifetime inequality is what you really care about, you will have to set targets in terms of thingswhich contributeto lifetime inequality:
• This suggests the need to focus on identifying states with long term or persistent effects, and then trying to affect them.
• It also suggests a focus on groups affected by persistentpoverty.
But it also implies a need to focus on recurrentpoverty. It is all very well to say that we can worry less about poverty if it istransitional, but that does not help very much: only small minority of low cross-sectional income is not associated with nearbyperiods of low income2. The implication is that while this kind of focus helps, most of the things associated with cross-sectional inequality are also associated with lifetime inequality and long term poverty. Therefore you should not give up onthe things you can measure.
The other part of the question in the title was about ‘increasing mobility’. Shouldpolicy focus on increasing mobility? And ifso, over what time scale:
• Short-term;
• Medium-term/life cycle;
• Intergenerational?
In other words, is mobility a ‘good thing’? A lot of the way people talk about it assumes that it is. Most people seem to be infavour of upwardmobility. We like the prospect of things getting better for ourselves and for our children. But we’re not sokeen on downwardmobility. People worry about how people will cope with a drop below accustomed living standards. At theextreme, some prefer an immobile society where “everyone knows their place” and so on, but even without such strong views,others may be squeamish about people having to adjust to substantially reduced circumstances.
If we think that a pound is worth less to the rich than to the poor, we are prepared to use redistributive taxation, but it is notobvious that policy would be counted a success if everyone simply swapped places with their opposite numbers in the incomedistribution each year. This has implications for measurement. It means that we probably want measures which are not simplyabout rank. That is, we may want to promote positive trajectories relative to, say average income, rather than simply measuringpositive trajectories in terms of movements between decile groups of the income distribution.
We are also worried about income risk, as mentioned before. Simple variability of income is not necessarily a good thing. Soit is only certain kindsof mobility or trajectories which we want. None the less if there is going to be some poverty, we mightprefer it was transitory, and so shared out as widely and briefly as possible, rather than concentrated on the same people. So indistributional terms we areinterested in income averaged over longer than just one week and so on, but at the same time highlyvariable incomes are effectively worth less than consistent ones.
But perhaps the most important reason why we are interested in mobility is that lackof mobility across lifetime or betweengenerations is a marker for lack of equality of opportunity. We may not be worried about intergenerational mobility per se, i.e.there is no particular gain from the wealth of parents being offset by the poverty of their children or vice versa. However, ifwe see that childhood poverty or growing up in a poor neighbourhood is associated with low education and lower laterearnings because of this, then we certainly should be worried.
So the answer to my original question is that it is hard to focus on lifetime inequality per se, but it does make sense for policyto focus additionally on things which have long-term effects – and hence have most purchase on life time inequality – as wellasfocussing on short term distribution. It also makes sense to focus on things which promote certain kindsof mobility, butagain not simply on promoting ‘mobility’ in aggregate.
The rest of the papers in this volume present a wide range of recent evidence on the factors and characteristics linked todifferent forms of mobility and immobility, crucial evidence in identifying policies with such effects.
1 The Dynamic of Welfare: The welfare state and the life cycle, edited by Jane Falkingham and John Hills (Harvester Wheatsheaf, 1995)2 See Stephen Jenkins’ contribution to this volume, or John Hills and Karen Gardiner, ‘Policy implications of new data on income mobility’ (Economic
Journal, February 1999).
3
‘Income dynamics in Britain 1991-6’1
Stephen P Jenkins, Institute for Social and Economic Research, University of Essex
1. Introduction
The aim of this chapter is to set out some of the basic facts about income dynamics in Britain during the first half of the 1990s.These descriptions provide some empirical background with which to help assess the normative issues raised by John Hills.Later chapters also do this and mine complements these in two principal ways. First, my focus is on household income ratherthan particular components of income and labour market earnings in particular (cf. Mark Stewart’s and Richard Dickens’schapters). And I look at the experience of the population as a whole, including both adults and children, rather than, say, onlymale workers. Second, I look at income changes from one year to the next. By contrast Robert Walker discusses lifecycletrajectories and Steve Machin documents intergenerational associations.
The data set used throughout is the British Household Panel Survey, waves 1-6 (covering 1991-1996). In section 2, I brieflyintroduce these data and give the definitions of key variables such as income. In section 3, I demonstrate that, although therehas been relative stability in the cross-section shape of the British income distribution during the 1990s, this disguises a largeamount of longitudinal flux at the individual level. I also provide some information about how much inequality in incomethere is if one takes account of this year-to-year income mobility, and how many people are touched by low income over aninterval of six years. Section 4 provides a first look at the factors which are associated with the movements in and out of lowincome. In particular it addresses the extent to which these people’s transitions are associated with changes in their household’smoney incomes, especially labour earnings, or changes in the composition of the household itself.
The research reported here is drawn from a longer paper (Jenkins, 1998), which provides many more details. Further analysisof the same topics, albeit based on data for BHPS waves 1-4, can be found in Jarvis and Jenkins (1997, 1998).
2. The Data and the Definition of Income
The British Household Panel Survey (BHPS)
The data summarised here are drawn from interview waves 1-6 of the British Household Panel Survey (BHPS), covering theperiod 1991-6. The first wave of the BHPS was designed as a nationally representative sample of the population of GreatBritain living in private households in 1991, and provided information about 10,000 adults in 5000 households. Subsequentlythese original sample respondents have been followed and they, and their co-residents, re-interviewed at approximately oneyear intervals in the Autumn of each year. Children are interviewed in their own right when they reach age 16. Six core topicareas are covered by the questionnaire each year: incomes, labour market activity, household demography and organisation,housing, health, and socio-economic and political values.
The definition of income
The income distributions used in this paper are defined in the same way as in the Department of Social Security’s HouseholdsBelow Average Income(HBAI) statistics: see for example, DSS (1998). These distributions have been specially created bySarah Jarvis and myself and are available from the Data Archive as a supplement to the BHPS release files.
More specifically the variable analysed is needs-adjusted household net income or simply ‘income’ for short:
‘Income’ = needs-adjusted household net income = household money income
household ‘needs’
where household money income is the sum over all household members of: head’s labour earnings, spouse’s labour earnings,other labour earnings, investment income, private and occupational pension income, benefit income, private transfer income(e.g. maintenance) minusincome taxes minuslocal taxes (currently the council tax). Household ‘needs’ are summarised by theMcClements before-housing-costs equivalence scale rate. For instance the needs ‘score’ for a childless married couple is 1.0and for a single householder, 0.61 (the scale rates also vary by children’s age).
In common with all other British household surveys, the income reference period is the month round about the time of theannual BHPS interview or the most recent period (except for labour earnings which refer to ‘usual’ earnings and investmentincome which is annual). In short, we have a measure of current rather than annual income. All income values have beenexpressed in pounds per week at January 1997 prices.
The unit of analysis is the individual (adult or child) rather than the household. Each person is imputed with the needs-adjusted net income of the household to which they belong.
1 Paper prepared for HM Treasury Workshop on “Persistent Poverty and Lifetime Inequality”, November 1998. The research reported here forms part of thescientific programme of the Institute for Social and Economic Research, which receives core funding support from the Economic and Social ResearchCouncil and the University of Essex.
Analysis of transitions into and out of low income is predicated on having a income cut-off in order to distinguish betweenhigh and low incomes. I define a person to have low income if his or her income is less than half average wave 1 (1991) income.
3. Longitudinal flux amidst cross-sectional stability
Table 1 provides a standard cross-sectional view on the income distribution for each year between 1991 and 1996. Onestriking feature is the stability in the shape of the distribution over this period. The Gini coefficient, a measure of the degree ofinequality, changed hardly at all, and nor did the proportion of the population with income below half the contemporaryaverage income. (This is of course in great contrast to the 1980s when inequality and low income rose a lot.) Along with thestability in distribution shape, there was some increase in real income, as shown by the rise in mean income by about one-tenthbetween 1991 and 1996, and the decrease in number of persons with incomes below the fixed half-1991 mean cut-off (equal toabout £130 per week).
There is a remarkable degree of longitudinal income mobility which co-exists with the cross-sectional stability in the shape ofthe income distribution: see Table 2. Each person has been classified into one of six income groups, from highest to lowest,according to their income ‘this year’ (wave t), and then classified again in the same way, but according to their income ‘lastyear’ (wave t-1). The table shows the outflow rates from each of last year’s income groups to the six different income groupsthis year.
There is a lot of mobility because, for all but one income groups, the proportion of persons in last year’s group who are in thesame group this year is about one-half or less. For example, only about one half (54%) of those with incomes below the lowincome cut-off last year also have incomes below the same cut-off this year. (The exceptional group is the very richest onewhere the stayer proportion is nearer three quarters, suggesting that income mobility is less amongst the richest.) On the otherhand, it is also clear that most year-to-year income mobility is short-range: observe the clustering of observations about themain diagonal of Table 2. For example, of those with incomes below the low income cut-off last year, 84% have incomesbelow 125% of the threshold income this year. This finding of ‘much mobility but mostly short-range’ is robust to the choiceof income group category and income. For example it is also found if one uses deciles rather than fixed real income values todefine the income groups (Jarvis and Jenkins, 1998).
The pattern of year-to-year mobility revealed in Table 2 (and other related work) can be summarised in terms of what I labelthe ‘rubber band’ model of income dynamics. One may think of each person’s income fluctuating about a relatively fixed‘longer-term average’. This value is a tether on the income scale to which people are attached by a rubber band. They maymove away from the tether from one year to the next, but not too far because of the band holding them. And they tend torebound back towards and around the tether over a period of several years. Of course over a longer period the position of thetethers will move with secular income growth or career developments. Also rubber bands will break sometimes if ‘stretched’too far by ‘shocks’—perhaps negative ones such as death of a breadwinner or positive ones such a big lottery win—leading tolong term changes in relative income position.
How much is inequality reduced if each person’s income is smoothed longitudinally?
Not very much: see Table 3. The statistics in the first column refer to incomes for just one year, 1991; the second column referto incomes longitudinally averaged over 1991 and 1992, and so on through to the last column, in which they refer to the six-year 1991-6 average. Inequality falls the more the income accounting period is extended—which is expected because theaveraging smooths fluctuations—but the decrease levels off relatively quickly. The reduction in the Gini coefficient from sixyear smoothing is roughly equivalent to the redistributive effect of direct taxes, as measured by the difference between the pre-and post-tax income Gini coefficients. The second row of the table expresses the same information differently. Standardisingthe Gini coefficients by averaged cross-sectional inequality provides a natural measure of income immobility: the moreincomes are longitudinally smoothed the less immobility there is. The statistic in the last column can be interpreted as sayingthat longer-term (six-year) income inequality is about 88% of inequality in an average year.
How many people experienced low income over a six year period?
Quite a lot, relative to the proportion poor in any single year (cf. Table 1). This is shown by Table 4. The first column showsthat more than two-thirds of persons in the sample never had a low income at any of the six annual interviews during 1991-6,and under two percent (1.7%) had a low income at every interview. On the other hand, the second column indicates thatalmost one third (32%) had low income at least once during this six year period and almost one fifth (19.1%) had low incomeat least twice. And just as inequality does not disappear with longer-term income smoothing, nor does the incidence of lowincome. If incomes are averaged over six waves, then the proportion of persons with smoothed income below the low incomethreshold is about 8%.
The upshot is that many more people are affected by low income over an interval of several years than have low income at apoint in time. This also means that the social security system helps a rather larger number of people than a focus on the currentcaseload would suggest.
4. What accounts for transitions in and out of low income from one year to the next?
I now turn to consider the factors associated with transitions into and out of low income, drawing on pioneering methodsemployed by Bane and Ellwood (1986). Recall that income (needs-adjusted household net income) is equal to household
4
money income divided by household ‘needs’. Thus changes in net incomes can arise via changes in household money incomes(‘income events’) and changes in household composition (‘demographic events’). I look at the correlates of low incometransitions by examining the relative frequency of different event types.
The analysis is based on a classification of event types according to a mutually-exclusive heirachy. This was done using thedefinitions summarised in Figure 1. Table 5 summarises the relative frequency of the different event types for transitions outof low income and transitions into low income. Not all low income transitions were included in the analysis: some wereexcluded to mitigate against the impact of potential measurement error (see the note to Table 5).
The breakdowns suggest that income events are more important than demographic events for accounting for low incometransitions, though demographic events are relatively more important for movements into low income than for movements out.More than four-fifths (82.3%) of transitions out of low income were income events of various kinds, which compares withabout three-fifths (62.4%) of transitions into low income. A person’s earnings can change because they get a different amountwhile staying in the same job, or because they get or lose a job altogether. Both are important. For example, amongst thepersons for whom a rise in household head’s labour earnings was the main event associated with a transition out of lowincome, the household head moved from not working to working in 51% of the cases. Amongst the persons for whom a fall inhousehold head’s labour earnings was the main event associated with a transition into low income, the household head movedfrom working to not working in 56% of the cases.
The breakdowns in Table 5 also draw attention to the relative importance of different types of income events. In particularthey highlight the relevance of earnings changes for persons other than the head of household. For example, for movementsout of low income, changes in the labour earnings of a spouse or other person in the household besides the household head arealmost as important as changes for the head. Non-labour income changes also play an important role.
There are of course some variations in the relative importance of different types of events when one focuses on differentsubgroups of the population. For example, demographic rather than income events account for the majority of the low incomeentry transitions amongst those currently in lone parent households. However the broad conclusions drawn in the lastparagraph are fairly robust across household types (Jenkins, 1998). The findings that demographic events and non-labourincome dynamics play significant role for many echoes those reported by Bane and Ellwood (1986) for the US during the1970s.
The results have an important implication for modelling income and poverty dynamics for the population as a whole. Evenwhere earnings events and low income transitions are relatively closely associated, we need to recognise that earningsdynamics are often a mixture of earnings dynamics for several persons (note role of spouse’s and others’ earnings). Acorollary of this is that a model of dynamics of a single income component (for example men’s wages) is unlikely to explainwell the income dynamics for any household type.
There are of course limitations to the Bane/Ellwood method used here. It is useful for identifying the salient facts to beexplained but does not tell us about causation or provide a means for dynamic simulation for policy analysis. One specificlimitation is that a hierarchical classification of events does not allow the unravelling of impact of simultaneous events. Moresophisticated analysis requires multivariate models. These are of two basic types, each with complementary advantages anddisadvantages (Jenkins, 1998). Top-down approaches include single equation models with low income status or income levelas the dependent variable: see the Oxley and Antonin chapter for an example of this approach. Bottom-up approaches modelthe underlying dynamic economic processes such as labour market participation and household formation and dissolution, andtheir relationships with various income components. The implications for income dynamics are derived from these buildingblocks. See Burgess and Propper (1998) for an application to the US. Given the extent of persistent low income and longer-term inequality in Britain, developing better models of both types is a research priority.
References
Burgess S. and Propper C. (1998). ‘An economic model of household income dynamics, with an application to poverty dynamics among American women’,Discussion Paper No. 1830, Centre for Economic Policy Research, London.Bane M.J. and Ellwood D.T. (1986). ‘Slipping into and out of poverty: the dynamics of spells’, Journal of Human Resources: 21, 1-23.Department of Social Security (1998). Households Below Average Income 1979-1996/97, Corporate Document Services, London.Jarvis S. and Jenkins S.P. (1997). ‘Low income dynamics in 1990s Britain’, Fiscal Studies 18: 1-20.Jarvis S. and Jenkins S.P. (1998). ‘How much income mobility is there in Britain?’, Economic Journal 108: 428-443.Jenkins S.P. (1998). ‘Modelling household income dynamics’, Working Paper 99-1, Institute for Social and Economic Research, University of Essex;Downloadable from http://www.iser.essex.ac.uk/pubs/workpaps/wp99-1.htm
5
Table 1Trends in mean income, inequality and low income: 1991-6
1991 1992 1993 1994 1995 1996
Mean income (£ p.w.) 259 269 272 274 288 290Gini coefficient 0.31 0.31 0.31 0.31 0.32 0.32Percentage below half contemporary mean 17.8 16.6 17.3 16.6 17.1 16.4Percentage below half 1991 mean 17.8 15.3 15.1 14.1 12.4 12.0
Number of persons (unweighted) 11634 11001 10473 10476 10119 10511
Source: BHPS waves 1-6, data weighted using cross-section enumerated individual weights. Income is needs-adjusted household net income per person inJanuary 1997 pounds per week.
Table 2Outflow rates (%) from last year’s income group origins
to this year’s income group destinations
Income group*, Income group*, wave twave t-1 < 0.5 0.5-0.75 0.75-1.0 1.0-1.25 1.25-1.5 ≥ 1.5 All (col. %)
< 0.5 54 30 9 4 2 2 100 (13)0.5-0.75 15 56 21 5 1 2 100 (22)0.75-1.0 5 19 48 20 5 3 100 (21)1.0-1.25 3 6 20 44 20 7 100 (16)1.25-1.5 2 3 8 25 35 27 100 (10)≥ 1.5 1 2 4 6 12 75 100 (18)
All 12 22 20 17 10 19 100 (100)
*: Income is needs-adjusted household net income per person in January 1997 pounds per week. Persons classified into income groups according to the size oftheir income relative to fixed real income cut-offs equal to 0.5, 0.75, 1.0, 1.25, and 1.5 times mean Wave 1 income = £259 per week. Transition rates areaverage rates from pooled BHPS data, waves 1-6, subsample of 6821 persons present at each wave.
Table 3Income mobility as reduction in inequality
when the income accounting period is extended
Income accounting period1991 1991-2 1991-3 1991-4 1991-5 1991-6
Gini coefficient 0.30 0.29 0.28 0.27 0.27 0.27Immobility index* 1.00 0.95 0.92 0.91 0.89 0.88
Source: BHPS waves 1-6, subsample of 6821 persons present at each wave. Income is needs-adjusted household net income per person in January 1997pounds per week. * Immobility index = Gini coefficient of income cumulated mperiods expressed relative to weighted average of Gini coefficients for incomein each of the mperiods.
6
7
Table 4The prevalence of low income
m Percentage of sample with low income at Percentage of sample with low income atm interviews out of 6 m or more interviews out of 6
0 68.0 100.01 12.9 32.02 8.0 19.13 3.9 11.14 2.7 7.25 2.8 4.56 1.7 1.7
Source: BHPS waves 1-6, subsample of 6821 persons present at each wave. Income is needs-adjusted household net income per person in January 1997pounds per week. The low income threshold is £130 pounds per week (half 1991 average income).
Table 5Movements out of and into low income broken down by type of event (column percentages)
Main event associated with low income transition Transitions out of low income Transitions into low income
Household head’s labour earnings change 33.6 31.0Spouse’s or other labour earnings change 28.5 15.9Non-labour income change 20.2 15.5Demographic event 17.7 37.7
All 100.0 100.0Number of persons 1684 1475
Income changes are money income rises for movements out of low income and money income falls for movements into low income. Analysis based on netincome transitions between one year and the following year, using pooled data for BHPS waves 1-6. Movements out of low income only included if netincome rose to at least 10% above the low income threshold; and movements into low income only included if net income fell to at least 10% below thethreshold. See earlier for definitions of income and low income threshold and of the hierarchy of event types.
8
Fig
ure
1
Cla
ssifi
catio
n of
‘inc
ome
even
ts’ a
nd ‘d
emog
raph
ic e
vent
s’ a
ssoc
iate
d w
ith a
low
inco
me
tran
sitio
nbe
twee
n la
st y
ear
and
this
yea
r (a
fter
Ban
e an
d E
llwoo
d, J
ourn
al o
f Hum
an R
esou
rces
, 198
6)
For
eac
h tr
ansi
tion
out o
f (o
r in
to)
low
inco
me
...
Hou
seho
ld h
ead
sam
e at
wav
es t-1
and t
Mon
ey in
com
e ch
ange
grea
ter
than
‘nee
ds’
chan
ge
Whi
ch in
com
e so
urce
incr
ease
d (d
ecre
ased
) th
em
ost?
‘Nee
ds’c
hang
e gr
eate
r th
anm
oney
inco
me
chan
ge
Whi
ch s
ort o
f dem
ogra
phic
cha
nge
is a
ssoc
iate
d w
ith th
e lo
w in
com
etr
ansi
tion?
Hou
seho
ld h
ead
chan
ged
betw
een
wav
e t-
1 an
d t
‘dem
ogra
phic
eve
nts’
Add
ition
s to
hou
seho
ld, e
.g. b
irth
of c
hild
, par
tner
ship
, oth
er jo
inin
g.Lo
sses
from
hou
seho
ld, e
.g. d
eath
of a
spo
use,
par
tner
ship
spl
it, c
hild
leav
ing
hom
e,ot
her
leav
ing.
‘inco
me
even
ts’
Cha
nges
in h
ead’
s la
bour
ear
ning
s, s
pous
e’s
labo
ur e
arni
ngs,
oth
er la
bour
ear
ning
s, n
on-
labo
ur in
com
e, e
tc.
‘Lifetime poverty dynamics’
Robert Walker, Centre for Research in Social Policy, University of Loughborough
Seebohm Rowntree recognised a lifetime dynamic in the poverty experienced by labourers at the turn of the 20th Century(Rowntree, 1901). They were born poor into a household with only one earner. After a brief period of comparative prosperity,prior to and immediately after marriage, the burden of child rearing took its toll, both adding to the demands on the familybudget and limiting the capacity to earn. Relative prosperity returned when the children began to earn and lasted after they lefthome, ending at such time as deteriorating health associated with increasing age reduced or eliminated the capacity to earn.
Table 1 Incidence of poverty in Britain, 1990’s
% poor in % persistently poor2 % permanently permanently 19911 (A) 1991-96 poor3 1991-96 (B) poor as a % of poor
Children 37 25 13 35
Single person without children 20 9 4 20
Couple without children 13 6 2 15
Single adult with child(ren) 64 53 29 45
Couple with child(ren) 29 17 8 28
Single pensioner 60 47 27 45
Pensioner couple 40 28 16 40
Source: Adapted from DSS based on the British Household Panel Survey.
1 Bottom 30 per cent of equivalent household income before housing costs.2 In bottom 30 per cent in four of the six years including the first.3 In bottom 30 per cent in all six years.
This temporal pattern in the incidence of poverty is evident today (O’Higgins, Bradshaw and Walker, 1988). Poverty1 is moreprevalent among children, families with children and the elderly than it is among other age groups (Table 1). Couples withchildren are well over twice as likely to be poor than those without, while single pensioners are three times as prone to povertyas single people under retirement age.
However, one cannot be certain that this pattern replicates the sequence of poverty and relative affluence described byRowntree. The reason is that, simulation studies apart (Falkingham and Hills, 1995), there are as yet no studies whichsystematically track the well-being of individuals from cradle to grave. It could be, for example, that the majority of peoplewho are poor in childhood are economically successful and escape poverty in later life, their places being taken by thedownwardly mobile who succumb to misfortune or indolence. While this may seem unlikely, it remains to be proved that theearly experience of poverty imprints itself indelibly on future life-chances, although it is known that people who have sufferedone spell of poverty are more likely than average to encounter another (Jarvis and Jenkins, 1997).
Despite the absence of adequate data, simply recognising the dynamics that may be inherent in poverty enhancesunderstanding of the phenomenon and opens the way towards better targeted policies. Time is not simply the medium inwhich poverty occurs, it forges its very nature. A single transient spell of poverty is not the same as permanent poverty(Walker, 1994). Likewise, the occasional spell has different personal, social and political ramifications from prolonged,recurrent periods below the poverty threshold (Ashworth, Hill and Walker, 1994). While poverty in childhood may not alwaysbe a significant social problem, poverty throughout childhood most definitely is. Even the possibility that poverty at one stagein the life course begets poverty at later ones demands investigation with the prospect of taking preventative action (Leiseringand Walker, 1998).
Making time explicit focuses attention on the three ‘T’s of poverty research:
• types of poverty;
• poverty trajectories;
• transitions and triggers.
Each is discussed below before briefly drawing some policy conclusions.
Types of poverty
Poverty has traditionally been conceptualised as both static and undifferentiated. A person has been counted as either poor ornot poor. It has even been rare to distinguish different degrees of severity (measured as the shortfall in income in relation to
9
1 Defined here, for convenience, as living in a household in the bottom third of the distribution of current income before housing costs.
the poverty threshold). However poverty should be differentiated and this can usefully be achieved in a number of ways. Twoare introduced below:
Temporal patterns
It may well be appropriate to differentiate poverty according to its temporal patterning. For example, simply classifyingpoverty according to the number, periodicity and duration of spells reveals much about the aetiology of poverty. Table 2summarises the experiences of the 38 per cent of Americans who experience poverty during childhood, while Figure 1illustrates certain of the different socio-demographic correlates associated with each type of poverty. Most children whoexperienced poverty in the 1970’s and 1980’s suffered repeated spells – 46 per cent recurrent lengthy spells or almostunrelieved chronic poverty – although 27 per cent experienced a single transient episode. Moreover, while 13 per cent of thechildren who were in poverty at any one time were likely to remain below the poverty threshold throughout their childhood,only five per cent of all children who ever suffered poverty were permanently poor.
Table 2 Types of childhood poverty in the US, 1968-1987
% of all children % of all children % of children who evercurrently poor experience poverty
during childhood
No poverty 62 – –
Transient1 10 27 5
Occasional2 3 8 4
Recurrent3 16 41 53
Persistent4 5 14 12
Chronic5 2 5 13
Permanent6 2 5 13
All 100 100 100
Children aged 0 to 15, poverty defined according to the PSID poverty threshold using equivalised annual income.1 Single one year spell.2 More than one spell, none lasting for more than one year.3 Repeated spells, some separated by more than one year, some exceeding one year.4 A single spell lasting between two and 13 years.5 Repeated spells never separated by more than one year.6 Continuous poverty.
Source: Walker 1994.
Figure 1 reveals that transient and occasional poverty were characteristic of poverty in the northern states of the USA whereaschronic and permanent poverty were largely confined to the South. Where permanent poverty existed in the North it wasentirely an urban phenomenon whereas in the South it was predominately rural. Not shown in the figure, 94 per cent oftransient spells were experienced by Caucasian American children, 76 per cent of permanent by Afro-Americans. All theCaucasian children afflicted by permanent poverty lived for a time with just one parent. Recurrent poverty, predominately theexperience of Caucasian children even though black children were disproportionately at risk, was associated with parents inlow paid jobs.
Figure 1 Childhood poverty in the US: type, region and rural and urban status of neighbourhood
10
Source: Walker (1994).
Transient Occasional
Persistent
Permanent
Chronic
Recurrent
39
2010
32
30
1030
30
5036
310
717
21
North-rural
South-urban
South-rural
North-urban
32
728
33
34
2534
7
Returning to Britain and to the incidence of poverty across the life course, Table 1 reveals that the ratio of permanent poverty –measured over six years – to the annual incidence of poverty is greatest among single pensioners and among lone parents. Notonly are these groups the most likely to be poor, they also confront the highest risk of being long term poor. On the otherhand, the poverty experienced by childless people of working age is disproportionately short-lived. Old age and child-rearing,especially in the absence of one parent, not only continue to increase the risk of poverty, in the way that Rowntree observed,they also help to shape the kind of poverty experienced.
Dimensions of poverty
Poverty can also be effectively differentiated by the various dimensions of well or ill-being (Table 3)2. Individuals’ locationwith respect to these dimensions may define whether someone is poor and the kind of poverty that they are currentlyexperiencing and may be likely to suffer in the future.
The ‘asset rich, income poor’ status of some older owner-occupiers has long been recognised (Walker and Hardman, 1988).Likewise, the income poverty of certain women and children has been documented within ostensibly affluent households(Millar and Glendinning, 1992; Middleton et al., 1997). Such women may also suffer a loss of dignity and autonomy. Indeed,some may even consciously choose another form of poverty: leaving home, they forego common property income and assets,preferring such autonomy as is possible living in refuges and bed and breakfast accommodation supported by benefit income(Leeming, Unell and Walker, 1994). In other circumstances, people may seek to avoid income poverty through copingstrategies, such as crime and prostitution, that serve to exclude them from society’s social and moral framework (Kempson,Bryson and Rowlingson, 1994).
In sum, there is not one kind of poverty but many.
Trajectories
Common sense, supported by a growing body of evidence, suggests that people’s future prospects are at least partlydetermined by their current location on the dimensions of well-being listed in Table 3. So, for example, someone who isunemployed and living on income-tested benefits but who has recent work experience, good qualifications and extensive socialnetworks is likely rapidly to find work that pays above benefit levels (McKay et al., 1997; Walker and Shaw, 1998). Theirchances of experiencing further spells of poverty, though greater than someone who has never suffered poverty, are alsocomparatively low and fall quickly as their spell of comparative affluence lengthens.
Alternatively someone in their late 50s, without skills or a fixed address, perhaps after spending years ‘on the road’, and whosuffers from chronic ill-health will probably remain trapped in poverty, finding a changed way of life difficult to secure orsustain (Vincent, Deacon and Walker, 1995).
Individuals follow trajectories through different kinds of poverty that may alternate with periods of relative affluence. Onedownward trajectory finds people moving from income poverty to social exclusion, slipping from a point of keeping theirheads above water, through ‘sinking’, to ‘drowning’ (Figure 2). This trajectory could be triggered by a breadwinner’s loss ofemployment and compounded by structural factors. The chances of finding work, and hence of escaping poverty, mightobjectively be low: few qualifications, obsolete skills, slack labour demand. This fact may be understood by the familyconcerned, undermining morale and hope that are both essential if work is to be found and financial ends met. The familymay, as a consequence, have no choice but to employ strategies at the socially unacceptable end of the list of available optionsand, hence, to step outside society’s norms.
Even before this point the family may have begun to become detached from social institutions, and the multi-faceted nature ofsocial exclusion become evident. Preventative health care may have been neglected. The demands of mutual reciprocity mayhave limited visits to all but closest kin, and stress and depression may have added impetus to this social retreat. Lack offinance may have prevented children from engaging in the full range of intra- and extra-curriculum educational activities withsignificant implications for their educational and social development. They may even have reacted negatively to inconsistenteconomic socialisation experienced in the home (Middleton et al., 1994). Confronted by the asocial behaviour to which thefamily has succumbed, the reaction of external agencies may confound attempts by the family to fight their social exclusion.Creditors may seek to repossess goods and landlords threaten eviction. Potential employers may be put-off by inadequatepersonal references and social welfare agencies may move to more coercive policies. The process of social exclusion will havebecome very hard to reverse.
Kempson et al. (1994) found that 12 out of 28 low-income families followed not dissimilar trajectories over a 24 month period.
Transitions and triggers
Progression along life course trajectories is punctuated by movement along the dimensions of well-being described in Table 3.These movements define transitions between levels of life quality, moves in and out of poverty, and between one kind ofpoverty and another. Such transitions are precipitated by triggers, the impact of which is mediated by interaction betweenindividual attributes, including a person’s location on the dimensions of well-being, and the opportunity sets determined bysocial institutions.
11
2 The dimensions are further discussed in Walker and Park (1998).
Tab
le 3
Elu
cida
ting
the
dim
ensi
ons
of p
over
ty
Labo
urIn
com
e1E
xpen
ditu
reS
tate
Ass
ets
Sha
red
prop
erty
Hum
an c
apita
lD
igni
tyA
uton
omy
mar
ket a
ctiv
itytr
ansf
ers2
right
s3
Em
ploy
edE
mpl
oym
ent
Spe
ndin
gS
ocia
l sec
urity
Sav
ings
With
inQ
ualif
icai
tons
Soc
ial s
tatu
sE
xter
nal l
ocus
full-
time
inco
me
paym
ents
hous
ehol
dof
con
trol
Em
ploy
edIn
tere
st/
Soc
ial
Inve
stm
ents
As
asse
tsE
xper
ienc
eS
ocia
l P
oliti
cal
part
-tim
edi
vide
nds
assi
stan
ceE
xclu
sion
excl
usio
n
Sel
f-em
ploy
edIn
kin
dC
once
ssio
nsB
orro
win
gsC
omm
unity
Hea
lth s
tatu
s
Out
side
labo
urH
ouse
(s)
Fre
e/su
bsid
ised
forc
e: c
ar s
ick
heal
thca
re
Con
sum
erF
ree/
subs
idis
eddu
rabl
esed
ucat
ion/
Veh
icle
(s)
tran
spor
t sys
tem
sP
olic
e
1C
orre
spon
ds to
priv
ate
cons
umpt
ion
in B
aulc
h’s
pyra
mid
.2R
elat
es to
cas
h tr
ansf
ers
thro
ugh
the
soci
al s
ecur
ity a
nd s
ocia
l ass
ista
nce
syst
ems
and
corr
espo
nds
to p
art o
f sta
te p
rovi
ded
com
mod
ities
in B
aulc
h’s
pyra
mid
.3
Cor
resp
onds
to
com
mon
pro
pert
y rig
hts
in B
aulc
h’s
pyra
mid
. I
n pe
rson
al c
orre
spon
denc
e B
aulc
h ex
plai
ns t
hat
in s
ocie
ties
of t
he S
outh
thi
s re
fers
to
reso
urce
s su
ch a
s la
nd h
eld
in c
omm
on.
He
was
uns
ure
wha
t m
ight
be
the
coun
terp
art i
n th
e N
orth
. A
dis
tinct
ion
is m
ade
here
bet
wee
n re
sour
ces
shar
ed w
ithin
a h
ouse
hold
and
com
mod
ities
pro
vide
d by
the
sta
te th
at a
re fr
ee a
t the
poi
nt o
f con
sum
ptio
n.
So
urc
e:
Wa
lke
r a
nd
Pa
rk (
19
98
) F
ig.
2.2
.
12
13
Fig
ure
2 A
traj
ecto
ry fr
om p
over
ty to
soc
ial i
nclu
sion
Tim
eA
ctiv
ityE
xpen
ditu
reIn
com
eA
sset
sT
rans
fers
Sha
red
Hum
an
Dig
nity
Aut
onom
ypo
vert
y rig
hts
capi
tal
T1
Loss
of
empl
oym
ent
T2
Ear
ning
s F
ewR
ecei
ves
low
Lo
wF
all
asse
tsbe
nefit
sM
oral
e
T3
Fal
ls n
ot
enou
gh
T4
Ado
pts
unac
cept
able
copi
ng s
trat
egie
s
T5
Wor
ks
Hea
lthR
etre
at fr
omH
igh
stre
ssill
icitl
yne
glec
ted
soci
alis
ing
T6
Inco
me
rises
Chi
ldre
n ne
glec
ted
T7
Dis
qual
ified
from
ben
efit
T8
Cea
ses
wor
kIn
com
e fa
lls m
ore
Soc
ial w
ork
invo
lvem
ent
T9
Exp
endi
ture
drop
s
T10
Thr
eate
ned
with
evic
tion
Key
: cau
sal l
ink
cont
inua
tion:
Sou
rce:
Wal
ker
and
Par
k (1
998)
, (F
ig. 2
.5)
Table 4 illustrates these processes. Thirty per cent of individuals who began a spell of income poverty in the early 1990’s didso coincident with the loss of employment by someone in their household.
Table 4 Events coincident with the beginning of a spell of poverty in Britain, 1990/01-1993/4
Household event % of persons who become % of persons experiencing poor and experience event event who become poor
Loss of earner (s) 30 27
Increase in number of earners 14 14
Loss of adult(s) 14 25
Addition of adult(s) 7 10
Loss of child(ren) 8 14
Addition of child(ren) 6 16Source: Walker and Park, 1998 (Table 2.2).
About one in seven of those becoming poor lost an adult as the result of divorce, separation, death or some other reason. Thesesame factors, together with the ending of entitlement to social insurance benefits, have been implicated as important triggers ofpoverty in other advanced western countries (Walker, 1997; Duncan et al., 1993).
However, neither the loss of a job or the departure of an adult from a household is either a necessary or a sufficient cause ofpoverty. About 73 per cent of individuals who either lost a job or lived in a household where someone else ceasedemployment escaped income poverty. So did 75 per cent of those experiencing the loss of an adult. These people must havebeen protected in some way from the worst consequences of events such as unemployment and relationship breakdown.
The source of such protection may on occasion be found among the other components of well-being. For example, povertymay be avoided in the event of unemployment by the provision of social security benefits. This is certainly the norm in muchof continental Europe. In Britain, though, even before Jobseeker’s Allowance further reduced the role of social insurance,insurance benefits prevented less than a fifth of unemployed persons from needing to claim social assistance (Walker andAshworth, 1998). Nevertheless, a tenth of workers who became unemployed during the early 1990’s recession made no claimon state benefits; some of these will have enjoyed the protection offered by redundancy pay, savings or a partner in paidemployment3.
Lifelong consequences
As already noted, it is not yet possible to document life-time trajectories nor to offer a complete dossier of the factors thateither provide protection against, or increase the likelihood of suffering, different kinds of poverty. Suffice to observe certaintransitions in later life.
Figure 3 suggests that the employment trajectories of older men may well be influenced by whether or not people have anoccupational pension. Whereas thirty per cent of men aged between 55 and 59 who were both working and belonged to anoccupational pension scheme had retired within five years, just 18 per cent of those who had no pension had done so. Theimplication is that men with occupational pensions may be able to avoid the income poverty that might otherwise be associatedwith early retirement and thereby choose to retire.
The same study (Tanner, 1998) reveals that very few men who were unemployed when in their late 50’s were back in workfive years later (12 per cent). However, while 40 per cent of those with pensions had retired this was true of only 25 per centof men without. Nevertheless, 82 per cent of the unemployed men without a pension left the labour market within five years,57 per cent moving into economic inactivity mostly supported by disability benefits. Men with pensions were much less likelyto follow this trajectory even when they had been unemployed in their late 50’s.
Figure 3 Transitions in late working life (ages 55-59)
No occupational {pension
Occupational {pension
Source: Tanner (1998).
14
3 Non take-up of benefits accounts for an unknown proportion of those not claiming benefits.
Figure 4 illustrates the pattern of protection afforded to women widowed in the early years of their partners’ retirement. (Itshould be noted in passing that the pension and investment income of men who died prematurely was only 67 per cent of thosewho survived (Johnson et al., 1998)). Typically widowhood caused a fall in income4, but one that was limited by pensionrights inherited from the dead partner (60 per cent of women received an average of £27 per week. A small number of women(eight per cent) also benefited greatly (£65 per week) from the NI widows’ benefits that are likely to be abolished after 2001(Cm. 4104, 1998).
Figure 4 The impact of widowhood in retirement
Source: Johnson et al. (1998).
Limited though this evidence is, it illustrates that the risk of income poverty can be mediated by the other components of well-being listed in Table 3. Moreover, the decisions that determined the risk of poverty associated with transitions in later lifewere often taken many years earlier.
Policy imperatives
Poverty comes in many guises with different aetiologies and varying immediate and, perhaps, lifetime consequences. Eachmay demand a targeted policy response. Passive social security measures may be appropriate in the case of transient spells ofpoverty but will typically not suffice in the event of recurrent or permanent poverty.
Research needs to identify, and policy to address, the factors that prevent triggers from precipitating episodes of poverty. Theaim must be to prevent the downward spirals described above and to stimulate upward trajectories in which the dimensions ofwell-being become mutually reinforcing.
At a time when family life and labour market circumstances may be increasing the risk of poverty (Beck, 1992; Leisering andWalker, 1998), individuals are being encouraged to take more personal financial responsibility for life-course contingencies:unemployment, ill-health and retirement (Cm. 3805, 1998). The quid pro quo is that government must help people to take onsuch responsibility:
• by offering timely and readily accessible advice ; and
• by providing or stimulating policy mechanisms that enable them to do so which are effective, cost efficient and risk free.
It is important that welfare provision should not become a gamble that further accentuates life’s lottery.
15
4 The decline in household income does not necessarily equate to a fall in living standard, which depends on the choice of equivalence scale.
Pre- Post-widowhood widowhood
(1988/9) (1994)
Other income
Survivor’s occupationalpension
Husband’s occupationalpension
Other state pension
NI widow’s benefits
NI retirement pension
References
Ashworth, K., Hill, M., and Walker, R., (1994), Patterns of childhood poverty: new challenges for policy, Journal of Policy Analysis and Management, 13, 4,pp658-680.Cm 3805. (1998) New Ambitions for Our Country: A new contract for welfare, London: Stationary Office.Cm 4104. (1998). ‘A New Contract for Welfare: Support in bereavement’, London. Stationary Office.DSS (1998) Households Below Average Income 1979-1996/97, London: Department of Social Security.Duncan, G. et al., (1993), ‘Poverty dynamics in eight countries’, Journal of Population Economics, 6, pp. 295-334.Falkingham, J., and Hills, J., (1995), The Dynamic of Welfare, London: Harvester Wheatsheaf.Jarvis, S. and Jenkins, S., (1996), Income Mobility in Britain 1991-94: A First Look, Colchester: ESRC Research Centre for Micro-Social Change, Mimeo.Johnson, P., Stears, G., and Webb, S., (1998) ‘The dynamics of incomes and occupational pensions after retirement’, Fiscal Studies, 19, 2, 197-215.Kempson, E., Bryson A., and Rowlingson, K., (1994), Hard Times, London: Policy Studies Institue.Leeming, A., Unell, J. and Walker, R., (1994), Lone mothers coping with the consequences of separation, London: HMSO, Department of Social SecurityResearch Report, 30.Leisering, L. and Walker, R., (1998) ‘Making the future: from dynamics to policy agendas’, pp265-286 in L. Leisering and R. Walker (eds), The Dynamics ofModern Society, Bristol: Policy Press.McKay, S., Walker, R., and Youngs, R., (1997), Unemployment and Jobseeking Before Jobseeker’s Allowance, London: Stationery Office, DSS ResearchReport 73.Middleton, S., Ashworth, K. and Walker, R., (eds), (1994), Family Fortunes: Pressures on parents and children in the 1990’s, pp. 159. London: CPAG.Millar, J., and Glendinning, C., (1992), Women and Poverty in Britain: the 1990’s, London: Harvester Wheatsheaf.O’Higgins, M. Bradshaw, J., and Walker, R., (1998), ‘Income distribution over the life-cycle’, pp. 227-255 in R. Walker and G. Parker (eds) Money Matters,London: Sage.Rowntree, S., (1901), Poverty: A Study of Town Lofe, London, Macmillan.Tanner, S., (1998) ‘The dynamics of male retirement’, Fiscal Studies, 19, 2, 175-196.Walker, R., (1994), Poverty Dynamics: Issues and Examples, Aldershot: Avebury.Walker, R. and Hardman, G., (1988), ‘The financial resources of the elderly or paying your own way in old age’, pp. 45-73 in S. Baldwin, G. Parker and R.Walker (eds) Social Security and Community Care, Aldershot: Avebury.Walker, R. and Park, J., (1998), ‘Unpicking poverty’, pp. 29-51 in C. Oppenheim (ed) An Inclusive Society, London: IPPR.
16
‘Childhood Disadvantage and Intergenerational Transmissions of Economic Status’
Stephen Machin1, Department of Economics, University College London and Centre for Economic Performance,London School of Economics
Introduction
Many people are interested in the extent to which inequalities persist across generations. It is straightforward to establish whyone should care about the extent of intergenerational transmission. For an offspring to be in an advantageous ordisadvantageous position simply because of their parents’ achievement has a distinct feel of unfairness to it, particularly froman equality of opportunity perspective. Many individuals across the political spectrum would champion the cause of equality ofopportunity and this is why accurately measuring the extent of intergenerational transmission is important. In the same way,pinning down the transmission mechanisms that underlie intergenerational transmissions is important, especially thoseassociated with childhood disadvantage.
Recent estimates of the extent of intergenerational transmission of economic status
Economists have typically considered intergenerational mobility in terms of earnings, income or education in two, rathersimple, ways. The first uses a tool commonly utilised in economics, regression analysis, whilst the second considersmovements up or down a distribution of interest. I therefore begin by summarising work on intergenerational earnings mobilitythat uses these approaches before turning to consider intergenerational transmissions of other measures of economic status.
The Regression Based Approach
The regression based approach typically specifies an earnings equation for members of family i of the form
(1) yichild = α + ß yi
parent+ ui
where y is earnings and u an error term.
In terms of equation (1) one can assess the extent of intergenerational mobility (or immobility) from estimates of ß: ß = 0implies complete mobility as child earnings are independent of those of their parents; ß = 1 implies complete immobility aschild earnings are fully determined by the parental earnings.
Most early studies in economics adopted this approach. The survey of this early work by Becker and Tomes (1986) states that ßwas generally estimated at around 0.2, leading them to conclude that “aside from families victimized by discrimination,regression to the mean in earnings in the United States and other rich countries appears to be rapid” [Becker and Tomes, 1986,p. S32]. However, more recent estimates have strongly challenged this view and pointed out serious methodological problemswith the early work (see Solon, 1992, Zimmerman, 1992, and Dearden, Machin and Reed, 1997, for more details on theseproblems). The following table summarises some of the more recent estimates, all of which show estimates of ß that tend to liesome way above the 0.2 “consensus” estimates described by Becker and Tomes. They all seem to imply a significant degree ofimmobility that violates the equality of opportunity characteristic of complete intergenerational mobility. For example, the‘typical’ father-son ß estimate in the Dearden, Machin and Reed (1997) study suggests that a son from a family (say family 1)with father’s earnings twice that of a father in another family (say family 2) earns 40-60% more than the son from family 2.
The Transition Matrix Approach
17
Author Data Estimate of ß
Becker and Tomes (1986) “Consensus” estimates from early (mainly US) studies About .200
Atkinson (1981) and UK data on 307 father-son pairs with sons subsequently traced 0.36 – 0.43Atkinson et al. (1983) (in the late 1970s) from 1950 Rowntree survey in York
Solon (1992) US panel data from the Panel Survey of Income Dynamics on about 0.39 – 0.53300 father-son pairs
Zimmerman (1992) US panel data from the National Longitudinal Survey of Youth on 0.25 – 0.54876 father-son pairs (but most estimates based on less than 300)
Dearden, Machin and UK panel data from the National Child Development Study (a cohort Sons: 0.4 – 0.6Reed (1997) of all children born in a week of March 1958) using earnings data Daughters: 0.5 – 0.7
for cohort members in 1991 and parents in 1974 – 1565 father-son pairs, 747 father-daughter pairs
1 This article is an updated version of my paper that appeared in CASEpaper 4: A B Atkinson and John Hills (editors) “Exclusion, Employment andOpportunity’, January 1998.
Of course the single number ß estimates given above are simply average estimates of the degree of intergenerational mobility.There may be important variations around this average. So the second commonly used approach for ascertaining the extent ofintergenerational mobility uses transition matrices which split the parental distribution of economic status into a certain numberof equal sized intervals (maybe quartiles, quintiles, or deciles) and then examines how many of their offspring remain in thesame interval or move elsewhere. An example, in terms of quartile transmissions (where one splits the parental earningsdistribution into four equal parts) based on data taken from the Dearden, Machin and Reed (1997) study is given below:
18
1565 Father-Son Son’s QuartilePairs
Father’s Quartile Bottom 2nd 3rd Top
Bottom .338 (.024) .297 (.023) .238 (.022) .128 (.017)
2nd .294 (.023) .312 (.023) .253 (.022) .140 (.018)
3rd .304 (.023) .243 (.022) .243 (.022) .209 (.021)
Top .064 (.012) .148 (.018) .266 (.022) .522 (.025)
747 Father-Daughter Daughter’s QuartilePairs
Father’s Quartile Bottom 2nd 3rd Top
Bottom .366 (.035) .321 (.034) .193 (.029) .118 (.024)
2nd .274 (.033) .305 (.034) .262 (.032) .160 (.027)
3rd .231 (.031) .219 (.030) .305 (.034) .246 (.032)
Top .129 (.025) .155 (.027) .241 (.031) .476 (.037)
The table shows, from looking at the leading diagonal, that the biggest proportion of sons who remain in the same quartile astheir fathers is in the top (i.e. highest earning) quartile. This is very marked with 52 percent of sons remaining in the topearnings quartile if their fathers were in that top quartile (for daughters the analogous percentage is 48 percent). The tabledemonstrates an important asymmetry in mobility with upward mobility from the bottom of the distribution being more likelythan downward mobility from the top.
Intergenerational Transmissions of Unemployment
Whilst most work on intergenerational mobility has looked at transmission in terms of earnings, income or educationalattainment, some work has looked at the unemployment status of sons and how it relates to unemployment experiences of theirfathers. In their analysis of National Child Development Study (NCDS) data Johnson and Reed (1996) report that 9.9 percentof sons had been unemployed for a year or more in the decade preceding 1991 (when they were aged 33). However, 19.1percent of sons whose fathers were unemployed at (child) age 16 experienced at least a year of unemployment between 1981and 1991.
Intergenerational Transmissions of Early Parenthood
Again using NCDS data Kiernan (1995) considers intergenerational transmission of teenage motherhood by looking at theextent to which young parents also had young parents themselves. A strong pattern is found with 26 percent of the cohort’steenage mothers also having a teenage mother, as compared to 10 percent of the cohort’s mothers who gave birth at the age of20 or after.
Summary
The research discussed in this section shows that an important part of an individual’s economic and social status is shaped bythe economic and social status of their parents. In the next section I go on to discuss work that tries to get into the ‘black box’of transmission to see what factors may underpin the strong intergenerational correlations depicted in the studies discussedabove.
Childhood disadvantage as a transmission mechanism
The principal impact of parents on their children is shaped in the childhood years of growing up. The most natural question toask is then: how important is childhood advantage or disadvantage as a transmission mechanism underpinningintergenerational mobility and how do they impinge on success or failure in economic and social terms in adulthood?
Ability in the Early Childhood Years and Parental Economic Status
In its strongest form (abstracting away from debates about genetic transmission) perfect mobility ought to suggest little relationbetween child ability and economic status. If a relation is uncovered one could think of this as being part of the transmissionmechanism underpinning transmissions of economic status across generations. The following regression (taken from Machin,1997) considers this question by relating the test scores of NCDS cohort members’ sons and daughters (aged 6-8) to parentalearnings (in 1991)2:
619 (Maths) / 617 (Reading) Child and Cohort Member Pairs (standard errors in brackets)
Child’s Maths Test Score Percentile = 6.932 ln(Parent’s Earnings)(1.939)
Child’s Reading Test Score Percentile = 4.720 ln(Parent’s Earnings)(1.903)
There is a strong relationship. A 50 percent higher level of log(parental earnings) suggests that a child would be around 3.5percentile points higher in the age 6-8 maths test score distribution and 2.5 percentile points higher in the reading scoredistribution. To the extent that these test scores are positively correlated with subsequent economic success (and quite a lot ofevidence says they are), then growing up in a family where father’s labour market earnings are high seems to be an importantstepping stone to having higher earnings later in life.
Childhood Disadvantage and Success or Failure in the Labour Market
So, how do early life factors like childhood poverty or social disadvantage influence individuals’ achievements in adulthood?In particular, how does growing up in a disadvantaged environment influence subsequent success or failure in the labourmarket? Gregg and Machin (1998) have considered this question in some detail by analysing data from the National ChildDevelopment Study (NCDS).
To understand how disadvantage transmits itself into adult life it is necessary to separate out the effects of childhood povertyfrom parental factors or innate child ability. Gregg and Machin do this by using the extremely rich NCDS data source3 tomodel economic and social outcomes in the earlier years of adulthood as a function of childrens’ development throughenvironmental, parental and individual specific factors. By following people through childhood and into the adult labourmarket this enables one to focus on the effects of factors like financial distress in the childhood years4 or measures of socialdislocation (e.g. contact with the police or truancy) after controlling for the early age ability of children (via test scores at age7) and other factors like parental education. The Gregg and Machin analysis of NCDS has uncovered important patterns thatdemonstrate a strong effect of childhood disadvantage on adult economic and social outcomes even once one nets out thesefactors.
At age 16 the main results are as follows:
• staying on at school, better school attendance and reduced contact with the police are more likely for children with higherage 7 maths and reading ability, for children with more educated parents and for children who grew up in families thatdid not face financial difficulties in the years in which children grew up;
• the impact of family financial difficulties is more important than family structure (whether the father was everunemployed, or living in a lone mother family);
• if children were ever placed in care during their childhood this massively increased their chances of contact with thepolice.
In terms of acting as a transmission mechanism underpinning intergenerational mobility, probably the key question concernsthe extent to which these factors impact on later economic and social success or failure. To investigate this, Gregg and Machinconsidered the relationship between economic and social outcomes at ages 23 and 33 and an array of measures of disadvantagein the childhood years.
Not unsurprisingly the educational attainment of the disadvantaged is considerably lower: for example, only 1 percent of boyswho had school attendance less than 75% or who had been in contact with the police went on to get a degree (or higher) by age23; this compares to 13 percent of the other NCDS boys. Figures for girls are 1 percent and 11 percent respectively. In terms offamily disadvantage only 4 percent of boys (3 percent of girls) who were ever placed in care or lived in a family facingfinancial difficulties went on to degree level as compared to 13 percent of boys (11 percent of girls) who were not in such asituation in their childhood years.
19
2 The regressions are based on children aged 6 years 0 months to 8 years 11 months at the time of the test and include a constant and controls for sex of thechild and the cohort member parent.
3 The NCDS data covers all individuals born in a week of March 1958 and the cohort members (and in some years their parents and schools) have so far beeninterviewed at ages 7, 16, 23 and 33 in 1965, 1974, 1981 and 1991.
4 The measure of financial distress used in this work closely corresponds to data on child poverty rates computed from Family Expenditure Survey data (seeGregg, Harkness and Machin, 1998).
At the other end of the education spectrum the disadvantaged are heavily over-represented in the part of the population thathave no educational qualifications. For example, 53 percent of boys (62 percent of girls) with school attendance less than 75%or who had been in contact with the police left school with no educational qualifications. This compares to 19 percent of boysand 25 percent of girls with better attendance and no police contact.
Because of these striking differences in educational attainment the research then considers whether the effect of disadvantageworks primarily through the fact that disadvantaged children have lower education levels or whether any effect still persistsonce one nets out the effect of education differences between those who are stylised as the disadvantaged and the otherchildren in the sample.
The age 23 economic and social outcomes looked at were: hourly wages; months spent in unemployment since age 16; whetherin a job at age 23; whether boys experienced any spell in prison or borstal since age 16; whether girls became lone parents by23. The age 33 outcomes looked at are wages and job status at 33.
The results demonstrate a strong link between poor economic and social outcomes at 23 or 33 and childhood disadvantage.Whilst part of the link is explained by the inferior education of the disadvantaged this is clearly not the whole story: even onceone nets out the effect of education differences the individuals characterised by disadvantage have significantly worse wages,unemployment time, employment and worse social outcomes (i.e. more likelihood of a prison spell for young men and of loneparenthood for young women). The key factors associated with disadvantage are both family based (growing up in a familyfacing financial difficulties, ever being placed in care in the childhood years) and child specific (low school attendance, contactwith the police).
All in all these results are strong evidence that childhood social disadvantage factors have an important impact on age 23 and33 outcomes. Even after netting out a variety of pre-labour market factors and educational attainment the less advantagedindividuals in the NCDS cohort are much less likely to be employed, to have experienced longer unemployment spells and/ordetrimental social experiences. Indices of childhood disadvantage like family poverty, family dislocations resulting in childrenbeing placed in care, poor school attendance and contact with the police seem to be important factors that underpin thetransmission of economic and social status across generations.
Trends in Child Poverty
One of the key findings of the Gregg and Machin work is that family financial distress during the childhood years displays animportant association with subsequent economic success or failure, and that this is a key factor underpinning intergenerationaltransmissions of economic status. This becomes of even more concern when one considers what has happened to child povertyrates in recent years.
Figure 1 based upon Family Expenditure Survey (FES) data and taken from Gregg, Harkness and Machin (1998), shows thatchild poverty rates have risen very fast in the last thirty years. Between 1968 and 1995/6 the FES data shows that, for incomeafter housing costs, the number of children living in poor households increased from 1.4 to 4.3 million (households are definedas poor if their equivalised income is below half the national average).5 The proportion of children living in poverty rose fromabout 1 in 10 in 1968 to just under 1 in 3 by 1995/6. Based on income before housing costs, the number of children in poorhouseholds rose from 1.1 million to 3.2 million (or from 8 to 24 percent) between 1968 and 1995/6. Both measures thereforeshow a striking increase, with child poverty tripling over this time period.
20
5 This includes all dependent children of school-age.
Note: The poverty line is defined as half average equivalised income in each year.
To the extent that the NCDS findings discussed above remain valid for more recent cohorts of individuals, these trends in childpoverty do not paint a promising picture for the evolution of economic inequalities in future. At the very least the results fromthis research suggest that child poverty should be on the policy agenda not only because of its increased incidence, but alsobecause of its potentially important impact on the future economic and social fortunes of adults who grow up in poorhouseholds.
Conclusions
Accurately measuring the extent of intergenerational mobility and understanding the factors that underpin such mobility orimmobility is important, especially when one bears in mind the rapid expansion in inequality that has occurred in Britain in thelast couple of decades. The policy implications of this rise become all the more important for future generations if there is notmuch mobility in economic status across generations. The results reported here suggest that, on the basis of study of quite largesamples of parents and children, the extent of mobility is limited in terms of earnings and education. Regression estimatespoint to an intergenerational mobility parameter (ß) of the order of .40 to .60 for men and .45 to .70 for women. Furthermore,from considering transition matrices, there is strong evidence of an asymmetry such that upward mobility from the bottom ofthe earnings distribution is more likely than downward mobility from the top. In the same vein, the early age cognitiveachievement of children is significantly related to the labour market earnings of their parents and to their parents’ maths andreading abilities. All this points to an important degree of persistence in economic success or failure across generations, centralto which is the ability of individuals to achieve higher earnings in the labour market.
Furthermore, factors associated with growing up seem to represent an important transmission mechanism that maintains thispersistence of economic success or failure across generations. Research based on the unique cohort data bases available in theUK shows that disadvantage in the childhood years has effects long into the adult life and there are often detrimental effectsthat spillover to the next generation. Having parents with low income or earnings during the years of growing up is a strongdisadvantage in terms of labour market success and can contribute importantly to factors like adult joblessness andparticipation in crime. The fact that these childhood disadvantages underpin the persistence of economic and social statureacross generations needs to be borne in mind by policymakers when designing policies that affect labour market outcomes inthe longer term.
ReferencesAtkinson, A. (1981). ‘On Intergenerational Income Mobility in Britain’, Journal of Post Keynesian Economics, 3, 194-218.Atkinson, A, Maynard, A. and Trinder, C. (1983), ‘Parents and Children: Incomes in Two Generations’, London, Heinemann.Becker, G. and Tomes, N. (1986), ‘Human Capital and the Rise and Fall of Families’, Journal of Labor Economics, 107, 123-150.Dearden, L., Machin, S. and Reed, H. (1997). Intergenerational Mobility in Britain, Economic Journal, 107, 47-64.Gregg, P., Harkness, S. and Machin, S. (1998). Child Development and Family Income, Forthcoming Joseph Rowntree Foundation report.Gregg, P., and Machin, S. (1998). Childhood disadvantage and success or failure in the youth labour market, Centre for Economic Performance DiscussionPaper 397.Johnson, P., and Reed, H. (1996). Intergnerational mobility among the rich and the poor: Results from the National Child Development Survey, Oxford Reviewof Economic Policy, 7, 127-42.Kiernan, K. (1995). Transition to parenthood. Young mothers, young fathers – associated factors and later life experiences, STICERD Welfare StateProgramme Discussion Paper 113.Machin, S. (1997). Intergenerational transmissions of economic status, in P. Gregg (ed.) Jobs, Wages and Poverty: Patterns of Persistence and Mobility in theNew Flexible Labour Market, Centre for Economic Performance Volume.Solon, G. (1989). Biases in the Estimation of Intergenerational Earnings Correlations. Review of Economics and Statistics, 71, 172-4.Solon, G. (1992). Intergenerational Income Mobility in the United States. American Economic Review, 82, 393-408.Zimmerman, D. (1992). Regression Toward Mediocrity in Economic Stature. American Economic Review, 82, 409-429.
21
‘Poverty dynamics in four OECD countries’
Howard Oxley, OECD1
Introduction
The OECD has recently examined poverty dynamics for four OECD countries (Canada, Germany, the United Kingdom and theUnited States) using longitudinal data sets which follow individuals over time and a common methodology. Longitudinal datasets enable flows into and out of poverty and the length of stay in poverty to be estimated, and allows identification of whichtypes of individual stay longest below the poverty threshold and whether certain changes in household status – such as gettingor losing a job or experiencing divorce – are associated with transitions into or out of poverty. The poverty line was defined asannual incomes falling below 50 per cent of the median of household disposable income adjusted for household size.
Because of differences in data sources, time periods and in the income concept used, one should be cautious about drawing toostrong conclusions from cross-country differences. Nonetheless, results point to a number of common features about povertydynamics across the four countries:
• Between 20 and just under 40 per cent of the population were poor at least one year over a six-year period. Within thisgroup, however, the majority had short spells;
• People experiencing longer spells have a lower probability of exit and the chances of exiting also fall with previousexperiences in poverty. At the same time, there is a high probability of falling back below the poverty threshold. Thus, forthose remaining poor over a longer period, low probability of exit and high probability of re-entry tend to reinforce each other;
• A small group of the population remains poor for longer periods of time with, apparently, very low chances of exit: peoplein poverty for six or more years typically make up 2 to 6 per cent of the population. However, because of their long stays inpoverty they represent from around one quarter to just over a third of the total time all individuals spend below the povertythreshold (from 30 to just over 50 per cent if five or more years are considered);
• The tax-and-transfer system sharply reduces the share of the poor, particularly among those poor for longer periods, andshortens the length of poverty spells;
• For three of the four countries, the characteristics of households experiencing shorter poverty spells tend to be differentfrom those with longer-term spells. Some groups appear to be over-represented among those experiencing longer-termpoverty, in particular women, lone parents and elderly single individuals, and the sick and disabled. A significant share ofthese have jobs;
• Obtaining or losing employment and improving earnings is particularly important for transitions across the povertythreshold. A large share of transitions occur when there are employment/earnings-related “events” (e.g. getting or losing ajob), particularly in the case of exits from poverty and this includes lone-parent households. Households with more than oneworker are better protected from becoming poor and have shorter poverty spells when they do. Increased employment orhours worked by other household members is an important source of poverty exit and households which get a second jobappear to shorten their spells by more than households which obtain a first job.
Poverty dynamics over a six-year period
Broad patterns of poverty dynamics
The poverty situation is both better and worse than the “static” poverty rates suggest (Figure 1). On the one hand, the share ofindividuals who are poor throughout the period is low (in the range of 2 to 6 per cent of the population); on the other, the shareof the population who were poor at least once over the six-year period is large (between 19.5 and 38.4 per cent of thepopulation). Indeed, the latter indicator is much higher than the “static” poverty rate, reflecting considerable turnover amongthose below the threshold2.
Table 1 (top panel) breaks down the share of individuals touchedby poverty over the six-year period on the basis of the number ofyears spent in poverty, including repeat spells3. In Germany, justover 46 per cent experienced poverty for only one year, which ishigher than for Canada, the United Kingdom and the UnitedStates (25 to 35 per cent). The opposite is the case for those poorfor five+ periods which make up 25 to 30 per cent of thosetouched by poverty for the latter two countries, compared withonly around 15 per cent for Canada and Germany.
22
1 Howard Oxley is a member of the Resource Allocation Division, Economics Department at the OECD. This work has been prepared by Howard Oxley,Pablo Antolin and Thai-Thai Dang with support from Ross Finnie and Roger Sceviour for the Canadian data. The opinions expressed in this report reflectthose of the authors and not necessarily those of the OECD.
2 The “static” poverty rate is calculated as the share of the poor in the total population in each year and then averaged over the period. The poverty rate is theshare of individuals in the total population who were poor in every year through the six-year period. The rate of those in poverty at least once is the share ofindividuals in the total population who were poor at least once through the period.
3 In the table, 5+ is the sum of five and six years or more spent in poverty.
An estimate of the share each of these group in the total time spent in poverty4 (bottom panel) indicates why the longer-term poorare central to policy: semi-permanent poverty is a more pressing social concern than transitory poverty because these groupsaccount for a large share of the total time spent in poverty; indeed, those in poverty at least five years make up as much as 50 percent (the United Kingdom and the United States) of the total time the whole population spent in poverty over the period.
Estimates of “hazard” rates for poverty exit (which indicate likelihood that people exit as a function of the time spent inpoverty) (Table 2, top panel) indicate that most individuals have short spells but that re-entry into poverty is also frequent(bottom panel). The decline in the “hazard” rates as the length of the spell increases indicates that the longer people remainpoor the more difficult time they have exiting5. On the other side of the coin, the longer the time since exit from poverty, thesmaller the likelihood of falling back in. Both patterns confirm recent research.
23
Table 1: Time spent in poverty over a six-year perioda
Share of individuals in poverty for 1 to 5+ yearsb (%)Post-tax and transfers
1 year 2 years 3 years 4 years 5+ years Average number of years in povertyCanada 1990-95 35.9 27.0 14.4 9.2 13.5 2.4Germany 1991-96 45.6 19.4 12.0 7.6 15.5 2.4United Kingdom 1991-96 26.0 19.3 13.6 13.2 27.9 3.1United States 1989-93 33.0 18.5 11.2 10.1 27.3 3.0
Share of total time spent in povertyc (%)Post-tax and transfers
1 year 2 years 3 years 4 years 5+ yearsCanada 1990-95 14.8 22.1 17.7 15.2 30.3 –Germany 1991-96 19.2 16.3 15.2 12.8 36.4 –United Kingdom 1991-96 8.3 12.3 13.0 16.9 49.6 –United States 1989-93 11.1 12.4 11.2 13.5 51.8 –a) The sample used includes all those individuals interviewed in each of the six years who have experienced at least one year in poverty.b) The share of individuals spending one to five+ years in poverty as a share of all individuals touched by poverty over the period. For example, 35.9 per
cent of those poor during the six-year period in Canada were poor for one year and 13.5 per cent for five years or more.c) The following steps were used to calculate the values in each column. First, the values in each of the columns of the top panel were multiplied (weighted)
by the number of years spent in poverty shown in the heading (distinguishing between five years and six+ years). A weight of six was given to groupswhich have six or more years in poverty, thus biasing downward the last column in the bottom panel. Second, these weighted values were then summed toestimate the total number of years spent in poverty by the total population. Finally, the values in the columns of the bottom panel are the results of the firststep divided by the total calculated in the second step.
Source: OECD Secretariat.
Table 2: Poverty dynamics: “hazard rates”a
Probability of exiting poverty afterb: (rate *100)Post-tax and transfers
1 year 2 years 3 years 4 yearsCanada 1990-95 55.7 41.3 38.8 35.4Germany 1991-96 52.7 42.7 32.0 19.1United Kingdom 1991-96 45.4 37.0 32.3 25.8United States 1989-93 45.6 31.9 23.1 20.2
Probability of re-entering poverty afterc: (rate *100)Post-tax and transfers
1 year 2 years 3 years 4 yearsCanada 1990-95 16.7 9.7 7.9 7.1Germany 1991-96 25.6 13.0 17.5 15.5United Kingdom 1991-96 32.8 18.2 11.0 10.0United States 1989-93 31.8 21.5 18.3 18.6
a) Latest available six-year period. Includes all poverty spells or spells above the threshold where the beginning of the spell can be observed.b) This is calculated as the ratio of those individuals, having just fallen below the threshold, who exit before the end of one, two, three or four years in
poverty relative to those still poor at the beginning of each successive period. For example, in Canada, 55.7 per cent leave before the end of the first yearand, of the 44.3 per cent who remain, 41.3 per cent leave between the first and second year.
c) The sample includes all those spells above the poverty threshold, conditional on the individual having exited poverty immediately before. The re-entryhazard is calculated as the ratio of those individuals observed to fall back below the threshold before one, two, three or four years above the poverty lineover the population at risk. For example, of those who moved above the threshold and are still above the poverty line after one year, 9.7 per cent will fallback below the threshold in Canada between first and second years.
Source: OECD Secretariat.
4 An estimate of the total time spent in poverty by group, obtained by weighting the shares in the columns of the top panel by the length of time each spendsbelow the poverty threshold (one to six years) and, then, divided by the total number of years spent in poverty by the whole population (the sum of theweighed values). (Table 1, top panel.)
5 This can reflect either a declining probability of exit because of feedback effects (such as the wastage of their human capital), or a sorting process in whichthose with the best chances of exiting leave first.
The characteristics of the poor by length of spell
A comparison of the family and labour-market characteristics depending on the time spent below the poverty threshold (Table3) (persons poor only one year and poor through the period compared with the non-poor), indicates a number of similaritiesacross countries although they do not apply to all countries in all cases6 (Table 4).
• The longer-term poor tend to be generally concentrated among those living in female-headed households, in lone-parenthouseholds, in single-adult households of retirement age7, in households where the head has lower education, and wherethere is no worker (Canada tends to be different). Nonetheless, a significant share of the longer-term poor work in allcountries, although this is less so in the United Kingdom;
• Again excepting Canada, the patterns for those poor only one year are less concentrated among households which areheaded by women, single adults, lone parents and the less educated than the longer-term poor and, in some dimensions,appear closer to the non-poor. Employment status differs significantly between the two groups;
• These results are supported by econometric tests of the factors affecting the length of stay in poverty (Oxley, Antolin andDang, forthcoming8).
Factors associated with poverty transitions
“Events” and transitions
Transitions into and out of poverty can be linked to certain employment or family-related “events” which can propelhouseholds to below the poverty threshold or permit them to rise above it. Although it is conceptually difficult to categorise“events” cleanly into one category or another – because family-related and employment-related “events” are often intertwined– the total number of transitions is broken down into three broad categories (Table 4):
• Transitions which occurred when there were employment/earnings-related “events” including changes in employmentstatus, hours and wage rates9;
• Transitions associated with family-structure-related “events”– mainly cases related to separation/divorce, partnerships/marriage, as well as children or other family members forming new households;
• Transitions with “other events” – which covers all transitions where there were no changes in either employment/earningsor family structure. These were mainly cases where there were large changes in transfer payments.
Focusing on the working-age population, the importance of employment/earnings-related “events” is high (although less so forCanada and the United Kingdom) and is even more marked for exits where they make up around 45 to 55 per cent (Canada,Germany, the United Kingdom) to over 60 per cent (the United States). In contrast, family-status-related “events” – whichsupplementary analysis shows are relatively more concentrated among poor single-adult and lone-parent households – aremore important for entries than for exits. Other events – which are largely transfer-income related – are more prevalent inCanada and the United Kingdom.
The frequencies shown in Table 4 do not show whether those experiencing an “event” are more likely to enter or exit poverty.To clarify this, further tests were carried out to see the effects of certain “events” on the chancesof entry and exit. To this end,Table 5 indicates by how much the probability of falling below or rising above the threshold changes if selected “events”occur.
Such tests of the correlation between “events” and transitions suggest that:
• For employment-related “events”, the probability of becoming poor increases the most where households lose all jobs orexperience a fall in hours, family situation held constant. In contrast, it is second workers (or an increase in hours in ahousehold with only one worker) which increase the probability of exit from poverty the most. For exits, a second earnerincreases the probability of exit by more than getting a first job10;
• The risk of poverty when an employment-related “event” occurs is lower if there is a stable household environment: theincreased probability of entry from separations and divorce rises sharply when they occur at the same time as employment-related changes.
24
6 Indeed, Canada appears to be quite different from the other three countries.7 Result based on supplementary information.8 Oxley, H., P. Antolin and T-T. Dang (forthcoming), “Poverty dynamics in selected OECD countries”, OECD Economics Department Working Papers.9 Certain cases where employment changes occurred at the same time as changes in household needs are also included in this sub-category. These tend to be
small in number and reallocating them to another category does not change the conclusions.10 It should be noted, however, that a significant share of all employment/earnings-related “events” shown in Table 4 concern workers who transit because
their earnings increase. This indicates that upward earnings mobility of those in jobs is also an important channel for exit.
25
Table 3: Characteristics of individuals above and below the poverty thresholdPer cent share of persons with a specified characteristic in each group
Household characteristics Total Above Poor for Always populationa thresholda 1 yearb poora
CANADAHead genderHead male 83.7 87.9 70.6 72.4Head female 16.3 12.1 29.4 27.6Employment statusNo worker 18.3 13.6 45.3 36.4One worker 31.2 27.9 42.3 43.9Two workers 39.0 44.3 11.3 17.9More than two workers 11.5 14.2 1.1 1.8Family typeSingle adult, no children 19.4 16.1 34.3 24.9Two adults, no children 30.1 32.4 15.5 26.7Single adult, children 4.4 2.1 11.3 11.6Two adults, children 31.5 32.4 33.2 29.2Large families 14.6 16.9 5.7 7.6Age of household headc
Young-age head 28.3 25.0 38.2 28.6Prime-age head 34.0 35.7 34.8 33.4Older-working-age head 21.8 22.6 21.6 22.0Retirement-age head 15.9 16.7 5.4 16.0Education leveld
Low education – – – –Middle education – – – –Higher education – – – –
GERMANYHead genderHead male 75.2 79.1 53.1 16.7Head female 24.8 20.9 46.9 83.3Employment statusNo worker 18.5 14.9 49.4 71.8One worker 39.4 37.6 44.1 28.2Two workers 34.8 39.4 2.5 0.0More than two workers 7.3 8.1 4.0 0.0Family typeSingle adult, no children 14.4 12.3 27.6 34.2Two adults, no children 40.9 43.2 31.7 14.9Single adult, children 2.8 1.3 15.5 32.1Two adults, children 33.6 34.7 19.0 16.7Large families 8.4 8.4 6.2 2.2Age of household headc
Young-age head 12.8 10.1 24.9 31.0Prime-age head 45.8 47.8 40.4 31.5Older-working-age head 26.6 27.7 17.2 10.7Retirement-age head 14.8 14.4 17.5 26.8Education leveld
Low education 28.0 26.0 29.6 30.5Middle education 52.3 52.7 60.9 63.0Higher education 19.7 21.4 9.5 6.6
UNITED KINGDOMHead genderHead male 67.7 74.3 53.7 38.4Head female 32.3 25.8 46.3 61.6Employment statusNo worker 25.0 10.7 42.7 91.0One worker 27.2 26.6 33.9 9.0Two workers 36.1 46.4 19.6 0.0More than two workers 11.7 16.3 3.8 0.0Family typeSingle adult, no children 10.9 6.2 18.1 40.5Two adults, no children 42.8 49.9 43.2 14.6Single adult, children 4.6 1.1 6.1 21.7Two adults, children 32.4 35.3 21.3 17.3Large families 9.2 7.6 11.3 6.0Age of household headc
Young-age head 15.5 13.3 17.6 23.5Prime-age head 47.7 53.6 39.9 23.4Older-working-age head 20.4 22.0 23.0 11.6Retirement-age head 16.4 11.1 19.2 41.5
26
Table 4: Frequency of “events” associated with poverty transitionsPer cent share of total transitions
Entries ExitsTotal population Household with Total population Household with
working-age heada working-age headaCanadaTransitions by typeb:Employment/earnings-related 26.1 28.1 38.4 44.4Family-structure-related 19.0 20.4 16.1 28.1Other factorsc 37.9 33.7 28.2 19.7
GermanyTransitions by typeb:Employment/earnings-related 47.1 51.5 50.1 55.0Family structure-related 24.7 25.2 8.9 9.5Other factorsc 23.4 17.9 33.3 27.1
United KingdomTransitions by typeb:Employment/earnings-related 27.5 34.2 41.2 51.6Family structure-related 26.4 28.5 9.1 10.2Other factorsc 35.4 23.4 40.5 26.6
United StatesTransitions by typeb:Employment/earnings-related 57.4 61.5 62.1 64.9Family structure-related 19.0 20.1 12.8 13.9Other factorsc 13.8 8.2 11.2 7.5Note: See Oxley, Antolin and Dang (forthcoming) for description. Covers most recent six-year period of available data.a) Refers to individuals in households with a working-age head.b) The per cent shares do not sum to 100 per cent. The remainder includes cases which could not be classified.c) Households which were either employed or unemployed in both periods and where income changes were not dominated by movements in earnings.
Largely transfer-related.Source: OECD Secretariat.
Table 3 (cont’d): Characteristics of individuals above and below the poverty thresholdPer cent share of persons with a specified characteristic in each group
Household characteristics Total Above Poor for Always populationa thresholda 1 yearb poora
UNITED KINGDOM cont’dEducation leveld
Low education 33.1 24.4 34.7 63.9Middle education 36.9 38.0 38.9 29.1Higher education 30.0 37.6 26.4 7.0
UNITED STATESHead genderHead male 79.5 86.8 70.4 26.9Head female 20.5 13.2 29.7 73.1Employment statusNo worker 10.7 6.6 17.9 51.3One worker 32.3 29.0 52.6 40.3Two workers 42.4 47.6 25.6 7.0More than two workers 14.6 16.8 3.9 1.4Family typeSingle adult, no children 12.3 10.6 20.0 21.5Two adults, no children 29.8 34.3 21.6 8.4Single adult, children 8.1 3.8 12.3 44.5Two adults, children 35.5 37.7 29.8 14.4Large families 14.3 13.6 16.4 11.3Age of household headc
Young-age head 19.7 16.0 33.3 37.1Prime-age head 51.7 55.1 41.6 35.5Older-working-age head 19.1 20.6 15.8 12.1Retirement-age head 9.4 8.3 9.3 15.3Education leveld
Low education 18.9 12.2 24.9 57.4Middle education 36.9 35.6 43.2 31.6Higher education 44.2 52.2 31.8 11.1
Note: Includes corrected values which differ from those found in Eonomic Outlook 64, Table VI.3. For definitions see Oxley, Antolin and Dang (op. cit.).Characteristics refer to the household head.
a) Characteristics defined at the beginning of the period.b) Individuals who are poor only one year over the period, excluding spells occurring in the first and last year of the six-year period.c) Young, prime-age, older-working-age, and retirement-age refer, respectively, to households with a head below 30, between 30 and below 50, between 50
and 65, and above 65 years old.d) Low education is less than high-school; middle is completed high-school and higher is more than high-school education.Source: OECD Secretariat.
Selected policy lessons
Briefly some main policy points are:
• Dynamic labour markets permitting good access to jobs are fundamental, permitting a rapid exit out of poverty for most ofthose falling below the threshold;
• In considering ways to reduce the number of poor individuals, governments may need to pay greater attention to thedistribution of earnings across households: reducing the number of no-earner households and increasing the number of two-earner households. Removing tax and labour-market disincentives to second earners may need to be considered;
• Specific help aimed at the longer-term poor is likely to be necessary as these groups often face a number of specificproblems which prevent labour-market entry;
• The fact that a fair share of the longer-term poor work suggests that systems to support low earners may need to beintroduced or strengthened. Careful design of these programmes is necessary.
27
Table 5: Changes in probability of entry and exit when an “event” occursChanges in probability points
Canadaa Germany United Kingdom United StatesPoverty entryb
Change in employment status onlyLoss of all workers 0.26 0.41 0.28 0.37Loss of some but not all workers 0.05 0.04 -0.01 0.08One worker, fall in hours – 0.15 0.12 0.22More than one worker, fall in hours – 0.00 -0.04 0.01
Change in family structure onlySeparations/divorce (spouse becomes head) 0.36 0.14 0.06 0.08Child becomes head – 0.09 0.09 0.16
Concomitant changes in employment and family statusSeparations/divorce (spouse becomes head)
Loss of all workers 0.72 0.82 0.49 0.62Loss of some but not all workers 0.47 0.31 0.05 0.25One worker, fall in hours – 0.58 0.28 0.45
Child becomes headLoss of all workers – 0.75 0.55 0.73Loss of some but not all workers – 0.23 0.07 0.38One worker, fall in hours – 0.48 0.34 0.59
Poverty exitb
Change in employment status onlyFrom zero to at least one worker 0.17 0.20 0.19 0.08Additional workers in working households 0.03 0.40 0.36 0.37One worker, increase in hours – 0.43 0.08 0.34More than one worker, increase in hours – – 0.30 0.47
Change in family structure onlyMarriage/partnership (head becomes spouse) 0.36 0.11 0.00 0.39Child becomes spouse – – 0.12 0.37
Concomitant changes in employment and family statusMarriage/partnership (head becomes spouse)
From zero to at least one worker 0.44 0.30 0.19 0.46Additional workers in working households 0.37 0.46 0.36 0.63One worker, increase in hours – 0.49 0.08 0.62
Child becomes headFrom zero to at least one worker – – 0.30 0.45Additional workers in working households – – 0.45 0.62One worker, increase in hours – – 0.20 0.61
a) Results for the Canadian data are less comparable to the other countries because information on labour-market and family status is more limited and someresults may be affected by breaks in the data.
b) Changes in probability points are defined relative to a baseline which is, essentially, the probability of transiting when there is no change either in familyor employment status.
Source: OECD Secretariat.
Session 1: Poverty and inequality from a dynamic perspective: Discussion
The main points from the presentations were:
• short term poverty is fairly widespread – lots of people could expect to experience short term poverty at some point in theirlives;
• but that persistent poverty is more concentrated;
• however, repeat spells of poverty are again fairly common;
• income mobility – there is a lot but most of it is short ranged;
• international comparisons are rather alarming – the UK has higher entry rates and lower exit rates;
• the evidence on inter-generational impacts was particularly startling – in particular the evidence that poverty and inequalityare not random events;
• the evidence that there are certain key events which make the experience of poverty much more likely – in particular losinga job and family change are significant factors in causing people to move into poverty.
The discussant raised the question of why we should adopt a dynamic approach when examining poverty. She argued that itcomes down to fairness. If poverty is concentrated and persistent we are more concerned than if it were random and shortlived. Also, if childhood poverty and inter-generational transmission has future effects we should be more concerned – becausewe are concerned about inequality of opportunity. And we want to understand the dynamics of poverty – to understand iscauses and how people move out of poverty, so government can devise policy to change people’s lives.
Issues raised in discussion included:
Measures of persitent poverty
If we are concerned about persistent poverty, why do we have no concerted policy measures against it? It is not simply down tolack of data – perhaps it is because the problem is too complex, or that it is easy to avoid the blame for persistent poverty.
The data
There were some arguments that it might be better to concentrate on simple sources of data ie this would give a clearer pictureof the problem. For example, there is lots of administrative data on lone parents, the long-term sick and disabled – groups weknow from elsewhere experience long-term poverty.
Focus on causes of success rather than failure
It was suggested that it might be more profitable to analyse what allows people to escape poverty to find out whatcharacteristics are needed – for example, housing tenure is particularly important (both as a cause and effect) – it is evident thatpeople are more likely to exit from poverty if they own their own home.
Arguments in favour of a dynamic approach
Cross-sectional data can provide interesting snap-shot analysis, but it is more useful in policy terms to analyse the same peopleover time. For example, people who are poor in their working lives will end up being poor pensioners – cross-sectional datawill not illustrate the mechanisms which cause people to live in poverty.
However, need to be careful when using data from the British Household Panel Survey for six years and calling it “long-term”– but it does give us a handle on causality which cross-sectional data does not.
Other issues raised:
• the importance of family events – one in six families having children move into poverty – would be useful to have moreinformation about this to develop policies to support family transitions;
• health inequalities – evidence which shows that the size of health inequalities is unrelated to income inequality and may infact be greater;
• the importance of inter-generational effects.
28
Session 2: Area and Multiple Deprivation
‘Patterns of area deprivation’
Sam Mason, Research, Analysis and Evaluation Division, Department of the Environment, Transport and Regions
Introduction
This paper first of all briefly discusses what are generally thought of as deprived areas, then goes on to describe DETR’s Indexof Local Deprivation (ILD) and the patterns of area deprivation shown by this. It finishes up with a description of the changesin patterns of area deprivation between 1991 and 1996.
What is a multiply deprived area?
Before looking at patterns of deprivation it is useful to try to sketch out what is generally thought of as a ‘deprived area’. Themain features of a deprived area are:
• a defined area which contains a high level or proportion of individuals or households who are experiencing a whole rangeof negative or undesirable circumstances, either singularly or in combination, which significantly reduce their overall wellbeing. e.g. low income, unemployment, crime, poor health, bad housing conditions and poor education;
• the concentration of these ‘deprived’ households and individuals in an area coupled with the undesirable aspects of the area– poor environment, poor housing, neglected open spaces, abandoned shops and houses, high crime levels, lack of services,lack of job opportunities to name but a few, act to reinforce the level of deprivation experienced by the community in thatarea;
• but it is important to recognise that most people in a deprived area will not experience all the factors of deprivation andsome people will suffer none of them. However, a significant number will experience many problems whilst many otherswill experience one or two problems. Also different groups of people will experience different combinations of problems;
It is also necessary to recognise that not all deprived people live in deprived areas:
• only 38% of all the unemployed lived in the 44 most deprived local authority districts as identified by the 1998 ILD and50% in the worst 65 LAs with similar figures evident for income support recipients.
Identifying multiply deprived areas
It has long been recognised that poverty and deprivation is more concentrated in some areas more than others and since the1960s policies have been developed with the objective of addressing the multiple problems of ‘deprived’ areas. In order toknow which areas these policies should be targeted on it has been necessary to develop measures which identify the areasexperiencing the highest levels of multiple deprivation.
As it is the concentration of many different problems which leads to area deprivation it is necessary to devise a measure thatcaptures this multiplicity rather than just single issues such as unemployment or low income. This has traditionally been donethrough indices of deprivation or disadvantage which combine a range of cross-sectional indicators into an overall‘deprivation’ score for an area and allow areas to be ranked relative to one another. Such indices have traditionally relied onsmall area data from the population censuses . DETR have produced indices of deprivation based on both the 1981 and 1991censuses which it has used to inform the targeting of its regeneration policies.
The DETR indices are only a small subset of the many different indices of deprivation, disadvantage and poverty that are inexistence. There are many others, all which have been designed with a slightly different purpose in mind and so all usedifferent indicators and different methods. For example, the Jarman Index was designed to help identify areas with the highestdemand for GP services whilst the Townsend Index was designed as an index of poverty. There is no single ‘right’ index,different ones will suit different purposes better than others. In general though, at the local authority district level it tends to bethe same authorities, albeit in a slightly a different order, that are consistently identified as the most deprived.
A further important feature of indices of area deprivation, DETR’s included, is they are measures of relative rather thanabsolute deprivation.
Deprived people or deprived areas?
As virtually all the data currently available for small areas is cross sectional, indices of area deprivation identify as manyindividuals or households in a particular area that are experiencing at least one of the selected deprivation factors rather thanthe number of individuals or households experiencing all or many of the problems simultaneously. So although an area canhave high levels on say eight different indicators of deprivation it doesn’t mean that any individuals or households in that area
29
will actually experience all those problems simultaneously. Indeed the principle of not using highly correlated indicators inindices of area deprivation means that the impact of individuals or households with multiple problems on the results arereduced in favour of those with just one problem.
Individuals or households experiencing multiple problems are obviously one of the most disadvantaged groups that social andeconomic welfare policy needs to address. Although we know from other types of analyses1 that such groups are concentratedin the most deprived areas, it is perhaps desirable to start thinking about how indices of deprivation can be made moresophisticated so that they more explicitly identify areas with high concentrations of people with multiple problems as well asmultiply deprived areas per se. The availability of data for small areas is slowly improving and in the forthcoming review ofthe ILD (see below) we will be considering how the concept of multiply deprived people in multiply deprived areas can bemore fully reflected in DETR’s Index of Deprivation.
DETR’s Indices of Local Deprivation
The Department of the Environment, Transport and the Regions’ latest index of multiple deprivation is known as the 1998Index of Local Deprivation (ILD). This had its origins in 1991 census data, when an index known as the Index of LocalConditions (ILC) was developed for what was then the Department of the Environment by Professor Brian Robson of theUniversity of Manchester.
The Index of Local Conditions
The ILC was produced at three spatial scales; local authority district, ward and enumeration district (ED). It contained 6indicators at the ED level and 7 at the ward, which were all taken from the 1991 census whilst a further 6 indicators from non-census sources were added at the district level to give a total of 13 indicators.
30
Included at all three levels:• Unemployment• Overcrowded households• Children in low earning households• No car households• Households lacking basic amenities• Children living in unsuitable accommodation (flats and temporary)
Included at the ward and LA District level only:• 17 year olds no longer in full time educationAll from the 1991 Census
Included at the LA District level only:• Income support recipients• 15 year olds with low grade GCSEs• Derelict land• Household insurance premiums (as a proxy for crime)• Standardised mortality rates• Male long term unemployment
The indicators were standardised, transformed and combined into a single score for each local authority district, ward and EDin England which could then be ranked relative to one another.
Spatial deprivation can manifest itself in many different ways and often its identification is highly dependent on the spatialscale at which it is measured. To help ensure that different spatial patterns of area deprivation were picked up the results of theILC were presented at local authority level in four different ways:
• the district ‘degree’ score – this is the LA district score based on all 13 indicators. It identifies the level of deprivation in thearea as a whole;
• the ward intensity score – this is the average index score of the worst 3 wards in the local authority and identifies areas withsevere pockets of deprivation;
• ward extent – this is based on the percentage of the LA’s population that lives within wards that are amongst the mostdeprived 10% in the country. Again, this identifies areas with pockets of deprivation;
• ED extent – this is the proportion of an LA’s EDs that are amongst the most deprived 7% of all EDs in England. Thismeasure identifies areas with small pockets of deprivation.
1 Analysis of the Sample of Anonymised Records from the 1991 Census of Population for DETR showed that 2⁄3 of children in Tower Hamlets, one of the mostdeprived boroughs on the ILD, were experiencing multiple problems in that they were living in no or low income households and at least two ofovercrowded households, households lacking amenities or households with no car.
Table 1
1998 Index of Local Deprivation
By 1997 the ILC was becoming increasingly out of date as it was based primarily on 1991 data. At the national level there hadbeen significant changes since 1991 in the numbers experiencing some of the problems, e.g. unemployment. This out ofdateness coupled with the Government’s increasing emphasis on directing policies towards those individuals and areas ingreatest need encouraged DETR to try and update its index of deprivation.
So, in February 1997 DETR recommissioned the University of Manchester to update the 6 non-census indicators in the districtlevel Index. As well as doing this, Professor Robson also recommended that some of the district level indicators from thecensus were replaced with more up to date substitute indicators that measured the same aspect of deprivation as the originalcensus indicators. He also recommended that one of the census indicators; children in unsuitable accommodation, was droppedfrom the index altogether as it wasn’t regarded a sensitive enough measure of the type of deprivation it was intended tocapture. For three of the census indicators it wasn’t possible to find suitable substitute indicators so the 1991 census data wasretained. In all it was possible to update or replace 9 of the indicators with data from around 1996. These, along with the 3retained census indicators gave a total of 12 indicators to make up the 1998 Index of Local Deprivation.
31
Updated indicators:• Income support recipients • 15 year olds with no or low grade GCSEs• Standardised mortality rates for under 75s• Non income support recipients of council tax benefit• Unemployment• Male long term unemployment• Derelict land• Household insurance premiums (as a proxy for crime)• Children in households receiving income support
1991 census indicators:• Overcrowded households• Households lacking basic amenities• 17 year olds no longer in full time education
There is very little data for small areas available from non-census sources (although this has got slightly better in the last fewyears) which meant it wasn’t possible to update the ward and ED level components of the index. So, in the 1998 ILD theseremain based on 1991 census data although the children in unsuitable accommodation indicator was dropped to maintainconsistency with the district level index.
Ward and ED Level:• Unemployment• Overcrowded households• Children in low earning households• No car households• Households lacking basic amenities
Ward level only:• 17 year olds no longer in full time educationAll from the 1991 Census
Patterns of Deprivation in 1996
Table 4 shows the 100 most deprived areas on the 1998 Index of Local Deprivation ‘district’ score. The most deprived areasare:
• the major cities and Inner London boroughs;
• closely followed by the metropolitan districts;
• the smaller industrial towns and cities;
• and some outer London boroughs.
Southern towns and cities, and resorts, start to appear after about rank 50 on the index e.g. Blackpool, Brighton and Hove,Norwich, Bristol, Thanet, Luton with the highest ranked rural authority being Penwith in Cornwall at 77. Resort and ruralareas tend to score more highly on the three pockets of deprivation measures than on the district level index as the smallerarea based favours the identification of deprivation in these areas, e.g. Blackpool is ranked 53 on the district level index but23 on the ward intensity measure.
Table 2
Table 3
In some regions high levels of deprivation cover large parts of the region: all of the local authorities in Merseyside are withinthe worst 50 in the country as are 11 of the 14 Inner London Boroughs. 15 out of the 18 districts identified as predominantlyport and industry (defined using the ONS classification of districts) are within the worst 50 and, 21 out of 25 coalfield districtsare within the worst 100 ranked authorities.
Table 4: 100 Most Deprived Local Authorities on the 1998 Index of Local Deprivation
32
Liverpool 40.07 1 Blackpool 20.14 51Newham 38.55 2 Easington 19.97 52Manchester 36.33 3 Tameside 19.78 53Hackney 35.21 4 Sefton 19.41 54Birmingham 34.67 5 Barrow-in-Furness 19.38 55Tower Hamlets 34.30 6 Leeds 19.06 56Sandwell 33.78 7 Westminster 19.05 57Southwark 33.74 8 Wansbeck 18.94 58Knowsley 33.69 9 Hounslow 18.89 59Islington 32.21 10 Brighton and Hove 18.75 60Greenwich 31.58 11 Wear Valley 18.67 61Lambeth 31.57 12 North Tyneside 18.67 62Haringey 31.53 13 Kensington and Chelsea 18.54 63Lewisham 29.44 14 Thanet 18.08 64Barking and Dagenham 28.69 15 Burnley 17.31 65Nottingham 28.44 16 Norwich 17.31 66Camden 28.23 17 Mansfield 17.30 67Hammersmith and Fulham 28.19 18 Preston 17.13 68Newcastle 27.95 19 Bristol 17.11 69Brent 26.95 20 Enfield 16.65 70Sunderland 26.90 21 Derby 16.37 71Waltham Forest 26.68 22 Luton 16.34 72Salford 26.64 23 North East Lincs. 16.20 73Middlesbrough 26.41 24 Wakefield 16.00 74
Sheffield 26.09 25 Portsmouth 15.86 75Hull 26.06 26 Hyndburn 15.84 76Wolverhampton 25.94 27 Penwith 15.78 77Bradford 25.94 28 Southampton 15.70 78Rochdale 25.13 29 Derwentside 15.26 79Wandsworth 25.05 30 Kirklees 15.23 80Walsall 25.02 31 Hastings 15.22 81Leicester 24.95 32 Great Yarmouth 14.72 82Oldham 24.82 33 Plymouth 13.86 83Halton 24.69 34 Harlow 13.50 84Gateshead 24.58 35 Wigan 13.47 85Ealing 24.48 36 Bolsover 13.42 86Hartlepool 23.72 37 Kerrier 13.32 87South Tyneside 23.67 38 Croydon 13.12 88Doncaster 23.60 39 Ipswich 12.80 89Coventry 23.48 40 Redbridge 12.80 90Blackburn with Darwen 23.04 41 Chesterfield 12.58 91Barnsley 22.30 42 The Wrekin 12.41 92Redcar and Cleveland 21.54 43 Ashfield 12.25 93Wirral 21.25 44 Blyth Valley 12.14 94St.Helens 20.98 45 Thurrock 12.11 95Lincoln 20.70 46 Calderdale 12.04 96Bolton 20.66 47 Torbay 11.86 97Stoke-on-Trent 20.61 48 IOW 11.85 98Stockton-on-Tees 20.41 49 Pendle 11.81 99Rotherham 20.23 50 Slough 11.75 100
However, as highlighted above it is also necessary to look at the ward and ED based measures when examining patterns ofdeprivation. Table 5 shows the 15 worst authorities on the 4 different measures and some interesting patterns are evident:
• Bradford and Leeds are ranked very high on the on ward intensity measure although they are much lower on the districtlevel index;
• Westminster scores highly on ward extent, ward intensity and ED extent measures but less highly on the overall districtscore;
• Middlesbrough and Hartlepool are ranked in the worst 15 on the ED extent measure but not on the other three;
• Barking and Dagenham scores highly on the district ‘degree’ measure but not on the ‘pockets’ measures.
33
Ward WardDegree Intensity Extent ED ExtentRank Rank Rank Rank
Liverpool 1 2 11 6Newham 2 10 2 3Manchester 3 4 8 8Hackney 4 8 1 1Birmingham 5 1 15 11Tower Hamlets 6 5 4 2Sandwell 7 29 19 40Southwark 8 14 6 4Knowsley 9 15 10 5Islington 10 22 3 10Greenwich 11 45 16 26Lambeth 12 9 5 7Haringey 13 12 7 9Lewisham 14 19 17 20Barking and Dagenham 15 65 22 89
Bradford 28 3 27 13Sheffield 25 6 36 32Leeds 56 7 40 49Kingston upon Hull 26 11 24 22Waltham Forest 22 13 12 25
Camden 17 32 9 23City of Westminster 57 20 13 21Hammersmith and Fulham 18 17 14 16
Middlesbrough 24 37 18 12Brent 20 16 21 14Hartlepool 37 47 25 15
Groups of deprived areas
One of the main uses of the ILD in DETR is to identify groups of most deprived local authority districts which are in turn usedto guide resource allocation. For example, 80% of Single Regeneration Budget Round 5 resources will be targeted on the 65most deprived authorities. These 65 authorities are those local authorities that are ranked within the top 50 on any one of thefour ILD measures. Analysis in the recent Social Exclusion Unit report ‘Bringing Britain Together’ used a similar approach toproduce a list of 44 ‘most deprived’ authorities (by taking the top 30 on any one measure) whilst the 17 local authority districtseligible to bid for New Deal for Communities pathfinder status were also selected on the basis of the ILD although a regionaldimension was also factored in. Because of it relativity there is no one way of using the Index to produce groups of deprivedareas. Different uses will require different methods.
Changes in Deprivation between 1991 and 1998
While the purpose of the ILD is as a measure of relative deprivation at a certain point in time it is also desirable to look atchanges in patterns of area deprivation.
Table 6 shows the fifteen most deprived local authorities on both the 1991 ILC and on the 1998 ILD. It can be seen that in
Table 5
general the places that were most deprived in 1991 remain the most deprived in 1998. In the 1998 index Barking andDagenham replaces Camden as the 15 the most deprived authority in England but the remaining 14 authorities stay the same,although there is some switching round in their rank position. Furthermore, although the ILC and ILD are very different to the1981 Index of Deprivation the same authorities were in general also ranked the highest in 1981.
34
1991 ILC 1998 ILD
Newham 1 Liverpool 1Southwark 2 Newham 2Hackney 3 Manchester 3Islington 4 Hackney 4Birmingham 5 Birmingham 5Liverpool 6 Tower Hamlets 6Tower Hamlets 7 Sandwell 7Lambeth 8 Southwark 8Sandwell 9 Knowsley 9Haringey 10 Islington 10Lewisham 11 Greenwich 11Knowsley 12 Lambeth 12Manchester 13 Haringey 13Greenwich 14 Lewisham 14Camden 15 Barking and Dagenham 15
However, due to the changes in the indicators and methodology used to produce the 1991 and 1998 indices, it isn’t strictlyaccurate to compare the 1991 ILC and the 1998 ILD. Therefore, as part of the contract to update the index Professor Robsonalso re-calculated a 1991 index using the 1998 indicators and methodology in order to examine the patterns of change betweenthe two dates. Changes in the availability and definition of data between the 1991 and 1996/8 meant that it wasn’t possible torecalculate an identical 1991 index. Even so, some tentative findings can be drawn on changes in relative area deprivationbetween 1991 and 1996.
Description of 1991-1996 change
At the top end of the scale the pattern is one of relative stability. Of the worst 50 local authorities in 1998 no fewer than 46 ofthese had been in the worst 50 in 1991 although again there was some switching around of positions. Lower down the rankingshowever, there is some evidence that certain types of area got worse between 1991 and 1996. Some geographically peripheralareas such as East Kent, West Cumbria and Coastal East Anglia, got considerably more deprived between 1991 and 1996 as didmuch of Outer London.
East Kent
• Dover was ranked 195 on the revised 1991 index but deteriorated to 103 in 1998;
• Thanet was ranked 76 in 1991 and rose to 64 in 1998;
• Hastings was ranked 110 and rose to 81, and
• Swale from 140 to 109.
Cumbria
• Allerdale was ranked 135 on the revised 1991 index and 102 on the 1998 ILD;
• Barrow-in-Furness was ranked 73 in 1991 and 55 in 1996.
Outer London
13 out of the 19 Outer London boroughs deteriorated between 1991 and 1996, many of them significantly:
• Hounslow jumped from 79 on the revised 1991 index to 59 on the 1998 ILD;
• Ealing went from 52 to 36;
• Brent went from 59 to 20 and was the LA which deteriorated the most between 1991 and 1996;
• Barking and Dagenham went from 35 to 15;
Table 6
• Croydon from 111 to 88;
• Enfield from 94 to 70.
But some of the most deteriorating authorities were still the worst LAs – Tower Hamlets, Southwark and Greenwich all gotrelatively worse over the period.
At the other extreme, some LAs did improve significantly between 1991 and 1996. Although there is no clear pattern to whichauthorities got better and which got worse many of the most improving authorities were northern towns and cities andmetropolitan districts (although it is necessary to bear in mind that they still remain some of the most deprived LAs in thecountry).
• Oldham improved from rank 11 in 1991 to rank 33 in 1996;
• Salford from 10 to 23;
• Bradford from 14 to 28;
• Leeds from 46 to 56;
• Barnsley from 32 to 42.
The major cities in the North; Manchester, Liverpool, Sheffield and Newcastle, remained in broadly similar positions, at themost only getting very slightly worse or very slightly better over the period. Liverpool was ranked the most deprived authorityon the recalculated 1991 Index and again on the 1996/8 ILD.
It should be stressed that these findings should be treated with caution as the changes may be the result of a change on just oneor two indicators. Likewise, it is necessary to recognise that the improvements and deteriorations are relative rather thanabsolute.
Small area deprivation
The construction of area-based indices is dependent upon the statistics that are available for small areas which, realistically arestandard administrative and census areas. This means that indices artificially define the boundaries of deprived areas. The ILDresults described above are all at a fairly large spatial scale and in reality area deprivation is likely to be much more unevenlydistributed across an area than these imply. Presenting the index in different ways obviously goes some way to capturing suchdifferences but up to date data at the ward and ED level is minimal which limits the effectiveness of these.
As it currently stands the ILD is most useful for identifying generally deprived local authority districts. This is sufficient forGovernment policy makers to use to draw up broad lists of the most deprived areas for targeting purposes, but is less useful foridentifying small deprived communities. Local authorities and other local institutions tend to have access to more and bettersmall area data for their own areas than is available nationally and are often is a better position to ‘map’ local level deprivationin their areas. It is therefore, important that these different sources are examined to obtain a full picture of deprivation in aparticular area.
ILD Review
Recently, however, more small area data has become available at the national level and it is hoped that these can beincorporated into the ILD so that it will be more reliable for identifying deprivation in smaller areas as well as at the macroscale. DETR are undertaking a further review of the ILD, of which one of the main aims is to try and improve the sub-districtlevel elements of the Index. The review will also look at the indicators that make up the ILD and the methodology used toconstruct it. It is hoped to commission an independent contractor to undertake the review by the end of December 1998 withthe results available in autumn 1999.
References1991 Deprivation Index: A review of Approaches and a Matrix of Results, HMSO, 19951998 Index of Local Deprivation: A Summary of Results, DETR, June 1998.Robson, B., Bradford, M., Tomlinson, R. Updating and Revising the Index of Local Deprivation. DETR, October 1998.
Further DetailsFurther information on the DETR Index of Local Deprivation is available on the DETR Website at www.regeneration.detr.gov.uk or from Dorrett Annon, Fax0171 890 3529, e-mail [email protected]
35
‘Area Problems and Multiple Deprivation’
Anne Power, Department of Social Policy and Centre for Analysis of Social Exclusion, London School of Economics
Introduction
Areas are made up of individuals grouped together, and like individuals they reflect a range of characteristics. These areacharacteristics both help determine and derive from the individuals who make up a local area or community. The physical areaand the residents of an area are inseparable parts of the same neighbourhood and lend each other identity.
Most areas have a mix of types of people and incomes within a range that gives them identity. Most areas have bothsignificant stability and mobility. Over a lifetime most people move several times within and across types of area. Areas act asboth a “sieve” and a “bucket” for people who move or stay. But some areas have a much narrower mix of people than averageand much higher mobility – Mayfair at one extreme, Spitalfields at the other. This discussion concentrates on areas of lowsocial or income mix at the bottom of the hierarchy. These areas tend to have high mobility.
Understanding area conditions and people
There are several questions: Do people at the bottom move around within poor areas, or do they gradually move up through ahierarchy of areas? What causes people to move up or stay at the bottom? Who moves into the spaces created at the bottom assome people filter out? Do the worst areas attract people with ever more concentrated problems leading to ever more marginalareas at the bottom? How can we measure an area’s position or area progress – through area change such as conditions, thevalue of property, the composition of population, or through the progress of individual people? Most area studies focus on areaconditions and the people who live there, rather than on individual progress within and through areas (Power, 1997). The mostbasic question is whether objective area conditions affect people directly or does the type of people living in an area determineits conditions?
In America, progress is measured in individual success rather than area change, with the consequent tolerance of appallingconditions in inner city ghettos, and many human casualties (Wilson, 1996). In Europe, including Britain, success is morecommonly measured by area improvement alongside individual progress. Arguably this is a more logical approach since thereis little doubt that areas affect people and people affect areas. The strongest and simplest proof of this lies in the cash valuethat attaches to near-identical properties in different areas. For property values are dictated by neighbours and neighbourhoods– in other words, area conditions.
In this brief discussion of problematic areas, I aim to show the interaction of ‘people’ and ‘area’ problems in acute urbandecline, the evidence of incipient abandonment in the most extreme areas, and the implications for policy initiatives such asNew Deal for Communities. The findings are based on CASE’s1 study of 12 high poverty areas and LSE Housing’s study forthe Joseph Rowntree Foundation, of incipient area abandonment in Manchester and Newcastle.
Concentrated poverty and clustering
The concentration of disadvantage in key regions, cities and neighbourhoods is significant. Taking wards (with an averagepopulation of 5,000) as the most easily measurable (but inexact) proxy for neighbourhoods, we found a large gap between the5% poorest wards and the average for England and Wales (Glennerster et al, 1999).
We defined poverty wardsas those within both the 5% of wards with highest proportion of people of working age not working,studying or training and the 5% with the highest concentration of deprived households (using the Breadline Britain Index).These two measures are not identical but together they identify the poorest areas in Britain (see Glennerster et al, for fullerdiscussion)
Table 1 showing % of population not working, studying or training and % in deprived* households in the poorestwards compared and England and Wales as a whole.
36
Poorest wards England & Wales
% of working age population not working, not studying,not on a government training scheme 45% 24%
% of households deprived* 38% 18%
*according to Breadline Britain Index
There are sharp regional differences in the concentrations of poverty as Table 2 shows. The extremely low figure for theEastern Region underlines the contrast between regions with the lowest and highest concentrations of deprivation.
1 Centre for Analysis of Social Exclusion.
Source:CASE Area Strand, 1998.
Even more stark is the high concentration of poverty wards in some local authorities and the clustering of poverty wards withinthese poorest local authorities forming large continuous areas of concentrated poverty. A poverty cluster describes an areawhere at least two high poverty wards are contiguous. Clustering is by definition an urban problem. Table 3 illustrates this:
Table 3 Local authorities with high % of population in poverty wards and large clusters of poverty wardsin continuous tracts
37
Merseyside 26%
North East 18%
North West 8%
Yorks and Humberside 6%
West Midlands 5%
Wales 5%
London 4%
Eastern 0.1%
Table 2 The percentage of the regional population living in poverty wards
% of LA Number of wards Population of population in in largest poverty largest poverty poverty wards cluster cluster in LA
1. Liverpool * 49% 26 259,000
Knowsley 54%
2. Manchester * 38% 16 175,000
3. Tower Hamlets * 57% 8 67,000
4. Middlesbrough * 46% 8 44,000
5. Bradford 8% 2 33,000
6. Hartlepool 39% 3 19,000
7. Hackney 31% 2 17,000
8. Rochdale 11% 2 16,000
Source: CASE Area Strand, 1998* Some local authorities have several clusters
Our analysis of the 1991 census identified 284 poverty wards i.e. wards that were both in the 5% most ‘work poor’ and 5%most deprived. Only 40 of these wards were ‘lone’ wards within a local authority. The rest were grouped in 51 clusters andadjacent areas.
Clusters of poverty matter because all the disadvantages associated with poverty are more concentrated and more extensivetherefore escape becomes more difficult. It is for this reason that large poverty areas persist, have a long history and attractpowerful stigma (DoE, 1976). The maps of deprivation, poverty and urban decline have largely not changed their place overlong periods (DETR, 1998). However, area conditionsdo change, are subject to decline, stabilisation or improvement,depending on what actions are taken (Robson, 1995).
Intrinsic and acquired area characteristics
The characteristics of areas that accumulate high poverty and deprivation can be divided between intrinsic or long-term areacharacteristics that make some places difficult and unattractive to live in; and acquiredcharacteristics that result from theimpact of intrinsic problems on people, determining who moves in, who stays and who moves out. Chart 1 summarises thesetwo categories.
Given the inequality of areas, for example distance from work, tenure, quality of schools and environment, it is inevitable thatmore vulnerable people with less economic clout will be concentrated in areas of greater difficulty, with lower opportunities.People with more choice will go to great lengths to stay out or move out of such areas. As a result, these areas have only halfthe proportion of people in work and double the proportion of deprived households.
The impact of clustering
The clustering of poverty neighbourhoods across large urban areas works to limit people’s chances in many ways:
• there are less obvious escape routes so more people feel trapped;
• depression and low morale are more common, resulting in lower levels of organisation and initiative and higher levels offrustration, aggression and other negative behaviour;
• children’s social learning is heavily influenced by surroundings and negative behaviour;
• schools suffer from low expectations resulting in lower performance and employment prospects; they also suffer moredisruptive behaviour;
• less cash income affects shops and other services;
• the high concentration of low-skilled people leads to intense competition for a shrinking pool of low-skill jobs, lower wagesand withdrawal from the labour market;
• the larger and longer running the area problems, the stronger the culmulative impact leading to flight of those more able togo and culmulative loss of control resulting from chronic instability;
• tipping into chaotic decline becomes more likely as the backbone of a neighbourhood weakens (James, 1995; Farrington,1996; Power and Tunstall, 1997).
These ‘clustering’ impacts on people’s life chances can be explained in the following way. At a personal level, being poor inan area with many poor people and poor conditions generates gradual loss of confidence in ‘the system’. In the largest povertycluster in Newcastle for example, only one in ten people vote. In Hackney and Liverpool the performance of local governmenthas been a source of scandal for two decades (DoE, 1979). Many conventional forms of involvement cease to operate as anarea declines. A sense of failure, rejection and shame over where people live and belong grows. This increases dissatisfactionand undermines hope of change.
All these pressures together generate aggressive behaviour particularly in young males (Power and Tunstall, 1997). Parentingand street behaviour often become ‘rougher’ under the impact of depression. A social climate can come to prevail in adepressed area that militates against collective provision, individual success and social cohesion. This then gives negativesignals to the next generation. The incidence of child abuse, youth crime, street disorder, disrupted classrooms, crime,shuttered shop-fronts and abandoned property stem from the loss of social controls that result from these forms of collectivewithdrawal. An area can slide from marginal viability towards collapse. This process was fully documented in Estates on theEdge (Power, 1997) and is supported by our current study in Manchester and Newcastle (Power and Mumford, 1999).
Evidence from declining cities and declining neighbourhoods
Evidence from four acutely declining neighbourhoods in two cities with severe long-run economic problems shows how severethe clustering of poverty is within large cities. The problems of these cities are different in degree, but not, I would argue, inkind from most other big cities. Intrinsic problems such as structural economic change can play into the more specific area –and people – based problems we have outlined. The four neighbourhoods we studied are within much larger poverty clusters.We present here only the barest summary of our findings to illustrate the severity of the problems we encountered. All tablesare derived from our forthcoming report for the Joseph Rowntree Foundation (Power and Mumford, 1999). I attempt to linkthese new findings to the wider problem of social exclusion and to the Government’s new approaches to regeneration.
38
Chart 1 Intrinsic and acquired area characteristics
Intrinsic characteristics Acquired characteristicsLocation Population mixTransport ReputationPhysical style Appearance/conditionsOwnership StandardsEnvironment Service performanceEconomy IncomeOutcomes OutcomesLow status Deteriorating conditionsLow value Rejection and isolationLow desirability Negative behaviourLow mix Withdrawal
There is long run movement away from cities and conurbations, although the exodus slowed over the 80s.
Source: Halsey, 1988; ONS, 1991.
With the exception of Inner London, the population decline is more acute in the main cities than the wider conurbations andeven worse in particular inner neighbourhoods. In the four neighbourhoods we studied, it is now leading to visibly abandonedpockets of sound housing.
Table 5: Depopulation of Manchester, Newcastle and four neighbourhoods
39
1961 1971 1981 1991
Greater Manchester 2720 2729 2595 2571
Tyne and Wear 1244 1217 1143 1114
Merseyside 1718 1657 1513 1412
Greater London 7992 7452 6696 6683
Table 4: Population of conurbations, 1961 – 91, ’000s
Period Manchester Newcastle *N1 N2 N3 N4
1971-1981 -18% -10% -39 -39 -13 -15
1981-1991 -11% -5% -5 -8 -19 -20
1991-1996 -7% -2% -6 -7 -20 -10
Source:Power and Mumford, 1999*N1-N4 represents the four inner neighbourhoods we studied in Manchester and NewcastleNote: from here on unless otherwise stated, the source for all tables is the report for the Joseph Rowntree Foundation by Power and Mumford, 1999.
The tenure pattern of the two cities is out of line with the national picture. A much higher proportion of local authority housingand far lower owner occupation help determine who lives and stays in the cities and who leaves. The pattern for the fourneighbourhoods is even more extreme.
Table 6: Housing tenure in 2 cities and 4 neighbourhoods
National Average Manchester Newcastle N1 N2 N3 N4
Local authority renting 19 38 35 50 54 48 77
Owner occupied 67 41 50 28 30 35 16
Private rented 10 12 9 8 8 10 2
Housing associations 4 7 5 13 6 6 4
The aspiration to own is difficult to meet within inner city areas and take-up of the right to buy is below the national averagefor both cities. In the four neighbourhoods, the level of right to buy is extremely low in spite of the predominance of goodquality council houses with gardens. People are less willing to risk ownership in the most acutely declining areas. Theconcentrated poverty affects aspirations and prevents many from considering it.
The loss of jobs in inner areas of both cities has affected males far more than females. The expansion of jobs in outer areashas not held people within the city boundaries as many people have leapfrogged to new housing in the surrounding districtswhile working in the city.
Table 7: Employment change in Manchester and Newcastle 1984 – 1991, as % of total working age population
All workers Manchester NewcastleInner –6% –7%Outer +41% +9%Male and female workers Male Female Male FemaleInner –13% +1% –19% +4%Outer +25% +54% –1% +14%
Source: DETR, 1996.
The cities, and particularly the four inner neighbourhoods, manifest acute signs of decline and deprivation. The contrasts withthe national picture are stark.
*as % of all unemployed
The intense deprivation of the four neighbourhoods affects almost entirely white populations. This confirms earlier findingsthat concentrated poverty in disadvantaged areas is frequently not connected with race. This is another important distinctionbetween British and American urban problems (Power and Tunstall, 1997).
In sum the most disadvantaged neighbourhoods experienced:
• a skewed tenure distribution with very large concentrations of social housing;
• significant and rapid depopulation over a long period and still continuing;
• ever more concentrated deprivation as in-comers tend to be more desperate than out-goers.
This resulted in:
• a rapid turnover of occupants and growing difficulty in keeping empty property filled;
• private withdrawal and growing empty spaces;
• trouble in the vacuum of collapsing demand;
• strong pressures to build outside the city so that people needing a home can escape the poorest inner areas. The processthus becomes self-fuelling.
These outcomes are illustrated by the following information we collected from small areas of incipient abandonment within thefour neighbourhoods.
Table 9: Turnover rate in Council housing as an indicator of low demand and severe management difficulty
40
National Manchester Newcastle N1 N2 N3 N4
% working age not working 24 31 37 46 48 49 50
% deprived 18 34 30 41 41 39 46
% long-term unemployed * 27 39 34 40 38 45 42
% manual 43 55 46 71 73 63 83
% children in lone parent households. 11 37 32 39 35 33 33
Table 8: Indicators of decline and deprivation in 2 cities and four neighbourhoods
Rate of turnover of tenants p.a.
National 10%
Manchester 18%
Newcastle 20%
Specific estates in four neighbourhoods (a) 40%
(b) 54%
(c) 30%
(d) 34%
(e) 36%
Table 10: Growing percentage of empty Council properties in 4 small areas
1996 1997 1998Area 1 2% 6% 19%
Area 2 7% 13% 16%
Area 3 6% 18% 35%
Area 4 4% 8% 15%
*Area 3 lost significant properties through demolition. This did not prevent continuing abandonment.
Table 12: Percentage of empty Housing Association and privately owned property
41
Table 11: Accelerated pace of abandonment of Council properties in one year 1997-98 showing growing number ofempty homes per quarter
Year start 1st quarter 2nd quarter 3rd quarterArea 1 75 111 136 166
Area 2 151 174 229 248
Area 3* 257 244 214 252
Area 4 249 360 370 430
% emptyHousing Association 1 54%
Housing Association 2 56%
Private landlord owned areas 1 42%
Private landlord owned areas 2 30%
Private landlord owned areas 3 50%
Incipient abandonment of unpopular areas
This summary of evidence points to a chronic problem that is relatively new in this country but long running in the UnitedStates – the tipping of areas from viability at the bottom of the hierarchy to galloping abandonment. The social consequencesfor the people left behind and for the surrounding areas which become affected by the blight are devastating. The knock-onconsequences for cities are worrying the government (Prescott, 1998). Large areas of major cities are losing value, cohesionand viability.
The following contentious questions are raised:
• Regeneration programmes, which are ongoing in these areas, have so far failed to stem the tide of decline. What changes ofapproach are necessary?
• Demolition of structurally sound and often physically attractive, renovated property appears inevitable in the face of zerodemand and zero market value. How does this fit with the overall projected increase in household numbers? Are thereuntried remedies to resolve this contradiction? Is it purely a regional problem?
• Private loft apartments and quayside flats within a mile of the emptying areas are selling vigorously for high prices. Coulddevelopers do more? Could less glamorous private initiatives help restore inner neighbourhoods?
• Since 1930 the allocation of council housing has been broadly needs-based. This has created intense polarisation andduring the 1980s it was made much worse by the loss of traditional jobs, the break-up of traditional family patterns, therapid expansion of owner occupation and the collapse of private renting. Is there a way of breaking up the pattern ofghettoisation?
• When social controls disintegrate under the impact of rapid change, crime and anti-social behaviour increase. How canorder and security be restored in such areas so that people with some prospects, commitment and confidence are willing tomove in? Is zero tolerance the answer?
• The trajectory for boys and men in poor neighbourhoods is strongly downwards. This is unleashing aggression andgenerating fear. Simply locking more people up is an expensive solution. What new ideas are there to tackle the alienationof many boys and young men?
• Residents within acutely declining areas face an increasingly precarious future. Some argue for new clearances and aclean-sweep, strategic approach to regeneration. Yet such solutions are immensely costly and therefore of limitedapplicability in the face of several thousand acutely declining neighbourhoods (SEU, 1998). Holding onto populations anddeveloping many smaller initiatives around them appears more promising. Will New Deal for Communities adopt theclean-sweep or incremental approach? What lower level initiatives are possible and practicable across every town and cityin Britain?
New approaches
Area based policies have two strong intrinsic merits. Firstly they operate within universal systems such as education, health,policing and housing that can link and equalise all individuals and communities in the country. Secondly they target specificadditional resources to compensate for the problems of accumulated and concentrated poverty, and to counter some of theintrinsic disadvantages of poor areas, the special negative effects of which are not in doubt. In addition the attention of
constantly renewed special programmes has a ‘Hawthorne effect’ – the extra effort and attention of themselves raise standards,at least for the duration of the programme. Most importantly, regeneration and other special programmes make theenforcement of basic standards within marginal areas essential. Programmes cannot succeed without this baseline.Regeneration programmes hold conditions and invariably arrest the slide into chaos that has struck many US inner cities. At itsbest it restores viability (Robson, 1995). A key finding from European experience of area reserve programmes in Estates onthe Edge (Power, 1997) is that even the most chaotically declining areas can be restored. Action gives out the strongest signalthat poor areas matter and poor people deserve equal treatment. For these reasons, special programmes targeted on specialneeds and poor areas are politically inescapable, logical, and a prerequisite for integration. Without them, in spite of universalsupports, some areas would decay even faster (Power, 1997).
So how can regeneration be made to last, to take root? In the new service-based economy of smaller and generally far better-off households, city authorities need to examine the potential for gentrification – attracting in and retaining many more of thenew-style workers who currently flee the city. This is happening among high earners in core city areas in highly secureddevelopments such as the quayside flats in Newcastle and canal-side loft apartments in Manchester. It could be made to workin the inner city hinterland by appealing to the following types of workers:
• low-income entrepreneurs and pioneers of new approaches;
• people who like the urban mix and value cities;
• commuters who prefer not to travel but are worried by schools, security, police and environment;
• people who see potential in extremely decayed, now obsolete inner areas.
Islington was a blighted slum clearance area until the 1970s. The central mill, warehouse and canal district in Manchester wasuntil the 1990s. The banks of the Tyne were until 3 years ago. There are many European and U.S examples of inner cityrenewal too (Urban Task Force, 1999).
To achieve such a shift towards a greater mix of population requires a new approach to regeneration. Crucially it needs pro-active visible policing, open access to good schools, intensive care of and quality of streets and open spaces, a broader group ofpeople in social housing, a change in ownership, the development of local shops and other services. These major policy areasthat are closely linked to the future of cities are beyond the scope of this short paper.
Many of the right ingredients are built into New Deal for Communities:
• a long-term approach;
• a focus on cities and distressed areas with the most severe problems;
• stabilisation and support for the existing population;
• the option of transfer of stock away from Council ownership.
But it must incorporate:
• open allocations systems for social housing;
• a revenue stream to fund the management of difficult areas of transition over the long-term;
• super-caretakers or wardens to perform the essential, front-line tasks of maintenance, security and social liaison;
• a new style of ground level guarding within neighbourhoods.
Overall it must take account of how to fund the organisational support task, how to generate on-going, smaller scalereinvestment, how to keep up with change while stabilising conditions, how to retain existing communities while attractingnew ones.
The Prime Minister’s proposals for super-caretakers and wardens within a framework of neighbourhood management at thelaunch of “Bringing Britain together” offer the potential for a changed approach to urban conditions (SEU, 1998).
Council housing needs to move in the direction of arm’s length, non-political, smaller scale structures, that diversifyownership, incomes and management, that have the freedom and flexibility to draw in new groups as old groups disappear.New housing company and housing association models are likely to spread rapidly. Already, Glasgow, Liverpool, Manchester,Coventry and Islington have discussed the idea of transferring all council housing away from direct political control.
42
Conclusion
Will the Government reverse the growing inequality of areas and stop the leeching out of more successful people? – only ifarea policy is linked to the wider urban agenda of re-concentrating cities, stopping sprawl, improving transport, raisingenvironmental and social conditions in cities as a whole.
My focus on small, deprived areas only suggests possible answers to the bigger questions of ‘city renaissance’ (Urban TaskForce, 1998). But successful cities and city neighbourhoods are an amalgam of a complex patchwork of initiatives, actions andenterprise within small localities (Jacobs, 1970; Power, 1997). By encouraging many of these patchwork initiatives in the mostdifficult areas alongside more ambitious and universalist reforms, the Government should at least reverse a slide intoabandonment, and at best lead on urban renaissance.
ReferencesCASE Area Strand (1998) Analysis of national census data to identify poverty wardsDETR (1998) 1998 Index of Local Deprivation. London: DETRDoE (1976) Inner Area Studies. London: DoEDoE (1979) Reports to the DoE on difficult to let housing in Hackney and Liverpool by the Priority Estates ProjectDoE (1981) Difficult to let investigation. London: DoEDoE (1996) Urban Trends in England: Latest evidence from the 1991 census. London: HMSOFarrington, D. (1996) Understanding and preventing youth crime. York: YPS/Joseph Rowntree FoundationGlennerster, H.; Lupton, R.; Noden, P. and Power, A. (1999) Poverty, Social Exclusion and Neighbourhood.CASEpaper, London School of Economics,forthcomingHalsey, A. H. (ed.) (1988) British Social Trends since 1900 – A guide to the changing social structure of Britain. London: Macmillan PressJacobs, J. (1970) The Death and Life of Great American Cities. London: Jonathan CapeJames, O. (1995) Juvenile violence in a winner-loser culture: socio economic and familial origins of the rise in violence against the person. London: FreeAssociation BooksJargowsky, P. (1997) Poverty and Place: ghettos, barrios and the American city. New York: Russell Sage FoundationONS (1990 & 1991)Labour Force Survey / Great Britain. London: ONSPower, A. (1997) Estates on the Edge. London: Macmillan PressPower, A. and Mumford, K. (1999) The Slow Death of Great Cities?York: JRFPower, A. and Tunstall, R. (1997) Dangerous Disorder: Riots and violent disturbances in thirteen areas of Britain, 1991-92. York: JRFPrescott, J. (1998) Press Release on launch of Urban Task Force. London: DETRRobson, B. (1995) Inner Cities Research Project: Assessing the impact of urban policy .London: HMSOSEU (1998) Bringing Britain together: a national strategy for urban renewal, Cm. 4045. London: TSOUrban Task Force (1998)Prospectus. London: DETRUrban Task Force (1999) Urban Renaissance – interim report. London: DETRWilson, W.J. (1996) When Work Disappears: the world of the new urban poor. New York: Hopf
43
‘Claimant dynamics of lone mothers – individual and area effects’
Mike Noble, Department of Applied Social Studies, Oxford University
This paper reports a study which examined the dynamics – or claiming patterns – of lone mothers receiving means testedbenefits. The study was undertaken in Oldham, a town in the North West. Although the focus was mainly on individualtrajectories on and off means tested benefits over time, the effect of ‘area’ was examined in two ways. First the fact that thestudy was undertaken in a single town effectively controlled for local labour market effects. Second we specifically examinedthe extent to which ‘area’ in the sense of local neighbourhood had any explanatory power when examining exits from IncomeSupport.
The study was based on two data sources. First, administrative data consisting of twice yearly extracts of all cases was drawnfrom the Housing Benefit/Council Tax Benefit data base for the study are beginning in July 1993 and ending in July 1997.These were then linked together to produce a longitudinal data set. Overall there were approximately 9,500 different loneparents represented within the data set over the study period. Second a postal survey of a random sample on lone mothers fromthe administrative data set was undertaken – to which just under 600 lone mothers responded.
The administrative data set
Table 1 Profile of Lone Mothers on Income Support July 1996
Mean IS claimants NS claimantsN=3123 N=1159
Age 32.5 33.8No. of dependent children 1.9 1.7Age of eldest child 9.4 10No. of children under 5 1.2 (1178) 1.1 (321)No. of hours working – 23.5 (760)Weekly Family Credit amount (£) – 60.4 (879)Weekly earnings (non self employed) (£) – 81.6 (911)Weekly self-employed earnings (£) – 52.5 (25)% of teenage mothers 1.5 0.6% under age 25 16.3 9% receiving disability benefits 1.9 5.3Tenure: owner occupiers 14 27.5
private tenants 31.3 21.8council tenants 54.7 50.6
% having children under 5 37.7 27.7% of earners receiving Family Credit – 94
The administrative data set covers both those on Income Support and those making a direct claim for Housing Benefit/CouncilTax Benefit as part of a package of ‘in work’ benefits. It includes the age and sex of the claimant, the number and age ofdependant children and housing tenure. It also contains limited information on hours worked, income sources and outgoingssuch as rent.
The administrative data showed that:
• the mean age of lone mothers on Income Support at any time during the observation period was around 32 years. Fewerthan 2 per cent were teenagers and 16 per cent were under 25;
• the mean number of children was approximately 1.9;
• the mean number of children under five was 1.2;
• the mean age of the eldest child was nine years five months;
• 55 per cent were council tenants, 31 per cent private and housing association tenants and the remaining 14 per cent owneroccupiers.
How much movement?
The results of the analysis showed a significant amount of movement on and off benefits revealing a variety of claimingpatterns among lone mothers.
44
At any one time point in the four year observation period there were between 2936 and 3398 (mean 3059) lone mothersclaiming Income Support – yet over the four year period a total of 6767 different lone mothers used Income Support for someperiod. This suggests a high turnover rate, with many lone mothers claiming for relatively short spells.
Chart 1: Origins and Destinations of Lone Mothers 1993-1997
Lone mothers who were on Income Support in July 1993 were tracked for the subsequent four years. Only 20 per cent had anuninterrupted claim for the whole period. 80 per cent had moved off benefit at some point. However, some of those whomoved off were unable to stay off and returned to Income Support. When movements off and then back on – ‘cycling’ – weredisregarded, 43 per cent of the lone mothers who were on Income Support at the beginning of the observation period were stillthere four years later (Chart 1).
An examination of non Income Support lone mothers (most of whom are in low paid work), revealed that only 32 per centwere still present on the data set as lone mothers four years on. Most had left altogether. 22 per cent were still present as nonIncome Support claimants and 10 per cent had become Income Support claimants.
From these results three claiming patterns were identified: long stay claimants, ‘cyclers’ – those who return to benefit morethan once – and short stay claimants.
Who moves?
Some of the characteristics of lone mothers associated with a greater propensity to stay on or move off benefit were identifiedby using techniques such as survival analysis and logistic regression. Several factors emerged as important.
• lone mothers under the age of 25 had the highest exit rate from Income Support. The results indicated that the older thelone mother, the less likely she would come off Income Support;
• the number of dependent children acted as a barrier to leaving Income Support. The more children a mother had, the lesslikely her exit from benefit became;
• in all analyses owner occupiers were much more likely to leave Income Support than those living in other accomodation.However, it was not possible to determine whether owner occupiers were simply better qualified or whether there weresome other factors at work such as where they lived.
The hypothesis that women who had their first child while still teenagers were more likely to remain on Income Support forlong periods was also tested. The findings clearly showed that:
• ‘Once teenage mothers’ were not significantly different from their older counterparts in their chances of leaving IncomeSupport and were no more likely to be long term reliant on Income Support.
The role of Family Credit
Charting the movements of Family Credit lone mothers from July 1994 to July 1997 revealed that:
• 30 per cent of the original lone mothers were still claiming Family Credit at the end of the period;
• 16 per cent were claiming Income Support;
• 54 per cent had either left the data altogether or moved into HB/CTB only claims which suggests that they had moved intohigher paid jobs not requiring the support of ‘in work’ benefits.
Thus although some lone mothers seemed to be reliant in the long term on a mixture of ‘in work’ benefits and low paid work, itwas a route out of benefit altogether for an increasing number as time progressed. There was however a small number whowere unable to sustain work even with the support on in work benefits and ended up on Income Support.
45
Time 1 (July 93)Not on data as lone mother
(2832)
IS(2936)
NS(984)
NS(1307)
IS(3398)
Time 9 (July 97)Not on data as lone mother
(2048)
95310
The role of area
To what extent does ‘area’ in the sense of the neighbourhood where a lone mother lives affect her chances of leaving IncomeSupport? If we map the quintiles of enumeration districts which have the greatest proportion of long stay Income Supportclaimants and compare these with the quintiles with the highest proportion of exits, on the face of it there does seem to besome clear, geographical distinction.
To investigate this further, cluster analysis was used to classify neighbourhoods into groups based on tenure, socio-economicand ethnic mix. These clusters were then incorporated into exit models. However, the effect such neighbourhoods had on exitsbecame insignificant as soon as tenure was controlled for.
When specific neighbourhoods were added into the models, one neighbourhood did stand out as having a significant effect onexits even when controlling for tenure. Investigation revealed that a cake factory with flexible employment policies waslocated in the area which helped to explain the higher probabilities of lone mothers leaving benefit in that area.
The sample survey
Though the administrative data can record gross movements on and off benefit and help identify some of the characteristics oflonger and shorter term claimants, it tells little or nothing of the reasons why, at an individual level, lone mothers move on oroff benefit. For instance, the administrative data cannot tell us if lone mothers came on to benefit as a result of relationshipbreakdown or whether they had always been single, never married and came on to benefit for other reasons such as the cost ofbringing up a child alone. Similarly, the administrative data yields no information on the reasons why, such as a newrelationship or a new job, lone mothers were able to leave benefit altogether. The postal survey gave information whichplugged some of these gaps in the administrative data.
The survey showed:
• most lone parents were married (41%) or living with a partner (21%) when they first had children;
• 30% described themselves as ‘single never married’ at this point, but only 22% consistently described themselves in thisway both when they first had children andat the time of the survey.
Reasons for coming on to and moving off benefits
The single never married lone mothers were typically in their mid to late ’20s and were most likely to have come on to IncomeSupport as a result of the birth of a child or difficulties combining work with childcare. Those who were divorced or separatedwere typically in their mid to late ’30s and were much more likely to have come on to Income Support as a result ofrelationship breakdown. The younger group of lone mothers tended to be better qualified overall, but this was still mainly atthe lower end of the qualifications spectrum. Work-related factors constituted the main reason for coming off Income Support,and in most cases this went with a successful claim for Family Credit.
Claiming patterns
The survey results confirm the variety of claiming patterns identified in the administrative data:
• estimates of length of time on Income Support showed that the majority had quite short spells on Income Support, but therewere a few very long stay claimants. About a quarter reported completed spells. The most common length of completedspells was two years. Long stay claimants were no more likely to be single never married or ‘once teenage mums’ thanother groups. Those who had not worked at all at any point and those without qualifications were more likely to be longstayers;
• 37% of lone parents who had ever been claimants of Income Support were ‘cyclers’ i.e. had multiple spells on IncomeSupport.
The overall pattern of results suggests that there are different groups of lone mothers with rather different family, work andbenefit ‘trajectories’. The single never married group tended to claim their Income Support when younger (during their ‘20s)and when they first have young children; the divorced or separated mothers were typically in their mid to late ‘30s.
Work orientation
All groups of lone mothers expressed a strong orientation to work. The majority had worked at some point since they first hadchildren. Although for many this was to ‘make ends meet’, there was a high positive endorsement of work for other reasonstoo. Only a very small number did not want to work more or at all. Younger mothers were more likely to want to work thanother groups (though they had worked less so far). Those who had not worked before or after children and did not want towork more or at all, were predominantly older mothers with married or long term relationship backgrounds.
Lone mothers overall held favourable attitudes to working for both positive (‘prefer to work’) and negative reasons (‘have towork’). Asked whether mothers with under fives should have to work, younger mothers had a very much more work
46
orientated position than older mothers (especially those 45+). This may reflect an attitude shift between lone mothers born inthe 1950’s and those in the 1970’s.
Younger mothers cited childcare reasons and financial consequences as the main reasons why they had not worked more.Older lone mothers were more likely to cite the need to stay at home to look after children or others in the family. Only aminority of lone mothers picked out personal (e.g. qualifications) or structural reasons (lack of available jobs) for not workingmore, though these would feature in any more comprehensive explanation.
Much of the childcare used by lone mothers was ‘informal’, though full time workers were much more likely to use formalcare than other groups. A high proportion of this informal care was provided by family and friends and was also paid for. Themain reason for utilising informal as opposed to formal care was the irregular and unsocial hours of much of the workundertaken by lone mothers. Informal care was the only feasible low cost solution. Lone mothers whose families lived locallyand were prepared to help with childcare were more likely to work. However, when lone mothers refer to ‘work’ or theirdesire to work, they do not necessarily mean full timework, they may have in mind part-time work with flexible hours that canfit around other needs such as those of their children.
Conclusions
From the survey it is clear that the overwhelming majority of lone mothers want to work and the analysis of the administrativedata demonstrate there is much movement into work.
For some, particularly those with young children or in the aftermath of relationship breakdown, work may not be the bestoption. It could be argued that mothers themselves should be the best judge of when and how much it is appropriate for themto work.
Many mothers who get into work find it difficult to sustain it both through problems with childcare and because of themarginal nature of the work they undertake.
This paper is based on Joseph Rowntree Foundation Findings April 1998 “Lone Mothers Moving In and Out of Benefit”.
47
Session 2: Area and multiple deprivation: Discussion
Evidence
Four issues were identified as fundamental to the debate on area and multiple deprivation.
• How important are area effects in determining outcomes for individuals?
Do we really see better outcomes for someone on a low income living in a well off area compared with someone with the sameincome in a deprived area? Evidence using the US Panel Study of Income Dynamics suggests not. However, studies ofexperiments in Chicago suggest that there are definite areas effects, with women who were moved into a better off area havinglower unemployment rates and children having higher school attainment than those still in the deprived area.
• Is there evidence of increased area concentration of poverty in recent years?
Data on this are limited, but work comparing the 1981 and 1991 census points to a drift towards more area concentratedpoverty, with increased polarisation into “rich” and “poor” areas with fewer areas left in the middle.
• Is policy too focussed on urban rather than rural deprivation?
The highest rural area on the Government’s Index of Local Deprivation is ranked 77 (out of 310). It was suggested thereforethat the index is biased towards urban problems, and the result is that rural areas do not get their fair share of regenerationfunding. The Index of Local Deprivation is currently being reviewed, and one aspect of the review will be seeing if it can bemade more sensitive to rural deprivation. However, the general point remains that any index of deprivation will pick up urbanfactors more.
• Do government indicators of deprivation measure the appropriate geographical areas?
Data used in the Index of Local Deprivation are collected for census Enumeration Districts, wards and local authority districts.These are historical geographical boundaries, and therefore in many cases it is correct to say that they may not representboundaries of local areas as they now exist. The Government is considering how more accurate data for small areas could bemade available.
Policy issues
“Mixed” areas
It was argued that there is a general need to create neighbourhoods that are more “mixed” in a number of different dimensions.The homogeneity of the population in deprived areas means a lack of enterprise, a lack of aspirations and a lack of money tosupport local services. One ethnic group often dominates, often resulting in problems of racism. Another point made here wasthat there is a need to break up the physical environment with different styles of housing and more open spaces. A moregeneral suggestion was that improving the perceived status of an area could be central to improving a wide range of aspects oflife there.
Social housing mechanisms
It was suggested that the social housing system was instrumental in creating and emphasising “poverty areas”, with local areasbecoming very homogenous in terms of workless or low incomes households. It was argued that this is largely because of thehomogeneity of these areas in terms of housing value and tenure, meaning that communities are made up of similar people.This was attributed to the hierarchical nature of the housing system, which means that people are always trying to move out ofareas lower down the scale, and no one on a higher income chooses to live there. A suggestion to solve this problem was thatlocal authorities should be allowed to sell off as much of the housing stock as possible. This could go some of the way tocreating more mixed areas, since good properties at low prices could attract a wider range of people into the areas. It wasclaimed that most of the social housing stock is in good condition (particularly outside London) and there are high levels ofunder-occupation in some areas. Therefore it was argued that the Government should not spend regeneration money onhousing, and should focus instead on improving the rest of the environment in these areas.
Points were made opposing the view that the social housing system was responsible for increased deprivation in the UK. Inparticular OECD work on deprived areas shows increased deprivation across all member countries, despite them having a widerange of different social housing systems. It was unreasonable therefore to attribute problems in the UK to the social housingmechanisms.
It was suggested that the UK should follow the French example of having non public ownership of housing, since havingcouncil landlords links the provision of social housing too closely with political objectives, building up the wrong incentivesand shifting the aim away from providing housing for those who need it.
48
Wider issues
It was suggested that we should look more widely at the factors impacting on deprived areas, most notably globalisation andconsumerisation. Both of these are putting pressure on any ideas of equity in the provision of housing. The result of these twoforces is and will be that people who can afford to live in better areas will almost invariably choose to do so, given that theywant a higher quality of life, better education for their children, etc. Hence society will become more polarised, with thesubsequent further increase in area based deprivation.
49
50
Session 3: Retirement and the Elderly
‘The decline of employment among older people in Britain’
Nigel Campbell
Older men have experienced the largest falls in employment over the last twenty years. Two-fifths of men aged between 55 and65 are now without work, compared to one-fifth in 1979. This paper1 uses data from the Labour Force Survey and the first sixwaves of the British Household Panel Survey to examine why older people’s employment has fallen, which groups have beenmost affected, and whether these trends are likely to continue.
Labour Force Survey evidence on aggregate employment rates
Chart 1 shows that employment fell substantially, for men of all ages, between 1979 and 1997 (similar points of the economiccycle). Older people have been particularly affected. 800,000 more men over 50 would be in work if employment rates had notfallen since 1979. This represents a larger fall in the number of jobs than the fall for “prime-aged” men (aged 25-50), eventhough the latter category is twice as large as the former.
Employment rates have always been lower among people in their 60s than in their 50s. Significantly, however, Chart 1 impliesthat there are now signs that the fall in male employment now begins from about age 50, rather than age 55.
Women are now much more likely to be in work than they used to be. Older women, however, have not shared in the generalrise in female employment. According to Chart 2, the largest increase has been among women in their late 20s and early 30s,while women over 55 are no more likely to be employed than they were in 1979.
The change in employment has mainly been offset by economic inactivity, with unemployment2 changing by much less. Inparticular, male economic inactivity has risen substantially as male employment has fallen, and female economic inactivity hasfallen dramatically as female employment rose. This is particularly true for older men – unemployment has not increased at allamong over 60s since 1979 – but is true at younger ages too. Men aged under 25, whose inactivity increased by slightly lessthan unemployment, are the only exception. Table 1 sets out the figures in detail.
1 This paper is an abridged version of Campbell (1999).2 These data are from the Labour Force Survey, and so use International Labour Organisation (ILO) definitions of labour force states. In order to meet the ILO
definition of unemployment, a person must have looked for work in the last four weeks and be available to start work in the next fortnight. Employment inthis paper includes both self-employment and working as an employee. People who are neither employed nor (on the ILO definition) unemployed, are said tobe “economically inactive”. The ILO definition is not the same as the claimant count of unemployment, which measures the number of people claimingJobseeker’s Allowance.
Table 1: Employment, unemployment and inactivity, 1979 and 1997 (%)4
Men Men Women WomenAll Aged 55-65 All Aged 55-60
Employment 1979 90.8 79.4 60.2 50.91997 80.6 58.3 68.9 50.4Difference –10.2 –21.2 +8.6 –0.5
Unemployment 1979 4.3 3.8 3.7 2.21997 6.2 4.6 3.9 2.2Difference +1.9 +0.9 +0.2 0.0
Economic inactivity 1979 4.9 16.8 36.0 46.91997 13.3 37.1 27.2 47.4Difference +8.4 +20.3 –8.8 0.5
Although the fall in male employment since 1979 is of a similar magnitude to the rise in female employment, there is no trade-off between them. There is no fixed number of jobs, as the (untrue) “lump of labour fallacy” would suggest. (See Campbell(1999) for more details of the arguments.) Measures to increase the number of older men in work therefore need not reduceemployment among younger people or among women.
Most of the fall in male employment since 1979 took place before 1983 (chart 3). However, this does not mean that thesechanges can simply be attributed to a one-off shock in the early 1980s. The data on unemployment and economic inactivity(chart 4) show that the picture is more complicated. One puzzle is why economic inactivity has continued to rise since 1993,while unemployment fell sharply, which contradicts the “discouraged worker” hypothesis3.
Regions with high male inactivity, at a particular point in time, also tend to have high male unemployment (chart 5). However,changes in male inactivity are not correlated with changes in male unemployment over the last cycle (1990-97). Charts 6a and6b show that, while the gap between regional unemployment rates has narrowed since 1990, the gap between regionalinactivity rates widened. Changes in male employment since 1990 are therefore not correlated with the level of regionalemployment then (chart 7). The three implications of these regional data are:
• It shows how drawing inferences about dynamic effects from “snapshot” data can be misleading;
51
3 This hypothesis suggests that “discouraged workers” (who would like work, but are not currently actively looking for it) are attracted back to the labourmarket when unemployment is falling. It implies that unemployment and economic inactivity should be positively correlated.
Chart 2: Female employment rates (%), by age
• It provides further evidence that the “discouraged worker” hypothesis is not a good explanation of what has happened to men inthe labour market since 1990. Either the hypothesis is flawed or a more powerful effect was working in the opposite direction;
• Even though the gap in regional male unemployment narrowed over the last economic cycle, there was no narrowing of thegap in total male employment. On the basis of the experience of the last cycle, economic recovery alone would seem not tobe sufficient to reverse the trend of falling male employment. That is not to say that the level of output does not affectemployment rates: deep recessions certainly harm employment, but recovery alone may not be enough.
This is an ongoing problem, not a one-off
We have already looked at the labour market status of people of different ages at two particular points in time (1979 and 1997),and at aggregate changes in between. Another way to analyse the Labour Force Survey is to compare people of the same agebut of different cohorts.4
52
4 A cohort is a group of people born in the same period. Here we consider cohorts of people born in the same five-year period, e.g. 1922-1926, 1937-31 etc.
Chart 3: Male employment rates (%)
Chart 4: Male inactivity and unemployment rates (%)
53
Chart 5: Regional unemployment and inactivity (%), 1997
Chart 6a: Regional changes in unemployment, 199097
Chart 6b: Regional changes in inactivity, 1990-97
One advantage of cohort analysis is that we can determine whether the decline in male employment:
• has particularly affected one or two cohorts; or
• is part of an ongoing trend, with each successive cohort of men more disadvantaged (in terms of lower employment rates)than previous ones.
Both these scenarios are consistent with the severe consequences, associated with the decline of traditional industries, for men whowere in their late forties or fifties at the start of the early 1980s. However, they have different implications for the future. Under thefirst, there is a clear concern for the generation affected, but the problem of declining employment would reduce naturally overtime. In particular, the employment rate would rise again once that generation reach retirement age. On the other hand, the secondscenario would give no grounds for complacency. It would imply that the trend is not likely to reverse of its own accord and, unlessthere were other offsetting effects, employment rates will continue to be significantly lower than they used to be.
The second scenario fits the data better. Later cohorts have lower male employment, at every age, according to Chart 8. Forexample, the employment rate of 50-year-old men born in 1942-1946 was lower than the employment rate of 50-year-old menborn between 1932 and 1936.
This implies that falling male employment is part of an ongoing trend, with each successive generation more disadvantagedthan the last, rather than a one-off effect that particularly affected one generation of relatively older men. The decliningemployment of older men is therefore not likely to reverse of its own accord. Extrapolating from the 1979-1997 data, and nottaking account of the effects of recent policy or other changes, we would expect to see the employment of older mencontinuing to decline (although not as quickly as in 1979-1983).
Chart 9 shows the opposite trend for women. Later generations of women are more likely to be in paid employment, reflectinggreater labour market participation. The increases from one cohort to the next are particularly large at younger ages. However,women are still less likely to be in paid work than men of the same age.
The wages of older workers
Real earnings for individuals continue to increase until a man (with the median wage for his age) reaches the age of 50 or so,and then decline slightly. However, at any given time, men around the age of 45 have the highest median hourly earnings.These statements can be reconciled as later cohorts have higher lifetime earnings and overtake those who are in their late 40sand are experiencing slower, but still positive, real earnings growth.
The average real wages of older men have increased over the last twenty years, but not by as much as the earnings of men intheir mid-40s. Older men have therefore seen both a disproportionate fall in employment and lower relative wages. Together,these provide evidence that older men have faced an adverse shock to labour demand.
32-year-olds have the highest median hourly wages among women. However, the decline in median earnings beyond that peakcan be fully explained by older women having on average lower educational qualifications.
Which individuals leave the labour market?
The Labour Force Survey is a representative sample of people at each age, so conclusions could be drawn from it about theemployment status of cohorts by comparing successive observations. As different people are interviewed every year, however,the LFS cannot tell us what is happening to particular individuals over a number of years. This is important because, forexample, the consequences of the following two situations may be very different, even if the overall employment rate at anygiven time is the same:
• people currently without work are likely to continue to be workless;
• people who are out of work today are likely to find work in the near future, so that individuals face shorter spells ofunemployment or economic inactivity. Even if the employment rate is the same as in the first scenario, more people wouldbe workless at some point in a five-year period, although the average durations of worklessness would be shorter. Inaddition, aggregate employment would be increased if more people had a close connection to the labour market.
Longitudinal data, which follow the same individuals over a period of time, are needed to investigate individuals’ transitionsbetween employment, unemployment and inactivity. The British Household Panel Survey (BHPS) can be used for this purpose,as the same people are interviewed every year.
This paper reports results from the first six waves of the BHPS. People who have given a full interview in all six waves haveprovided information on their labour force status on 1st September in seven years (1990 to 1996 inclusive), and whether theyhave had any spells on unemployment or long-term sickness during these years.5
54
5 We also include those for whom we have three or more observations of whether or not they were employed, but who were not interested in every year of thesample . Paull (1996) has demonstrated that excluding everyone who would not be tracked down for an interview in every wave of the survey will givebiased results. In total, the sample contains 2253 individuals aged between 45 and state pension age when their labour force status was first observed. Wehave a complete record – from September 1990 to September 1996 – for 73% of those in the sample.
55
Chart 7: Regional changes in employment, 1990-97
Chart 8: Male employment rate for different cohorts
Chart 9: Female unemployment rates for different cohorts
These data are used to provide evidence of who leaves the labour market, and who returns.
Who leaves the labour market?
Of the 71 per cent of the sample who were employed initially, 56 per cent of these were employed throughout their period ofobservation (typically six years), while 44 per cent were displaced from the labour market at some point. The latter may bebecause of a period of unemployment, long-term sickness, or (voluntary or involuntary) retirement. People who move from jobto job without a period of unemployment or long-term sickness are held to have been in continuous employment.
Table 2 reports the results of a probit6 regression to determine which characteristics are associated with whether or not a personis displaced from the labour market. It gives the probabilities of someone being displaced during the sample period (typically1990-1996) given their age, gender, wage quartile, whether or not the person was working in a declining industry, andoccupational pension scheme membership.
Table 2: Probability of being displaced during the BHPS sample period (%)
Men Women
Age at beginning 45-49 50-54 55-59 60-64 45-49 50-54 55-59
In non-declining industries:Q1 of wage distribution and OP 19.6 28.9 39.6 65.6 25.6 36.1 52.5Q1 and no OP 12.1 19.3 28.2 53.6 16.6 25.2 40.2Q2 and OP 17.8 26.7 37.0 63.1 23.4 33.6 49.8Q2 and no OP 12.1 19.3 28.2 53.6 16.6 25.2 40.2Q3 (with or without OP) 16.9 25.5 35.7 61.8 22.4 32.3 48.4Q4 (with or without OP) 20.6 30.2 41.0 66.9 26.7 37.5 54.0
In declining industries:Q1 of wage distribution and OP 29.3 40.4 51.9 76.2 36.5 48.2 64.6Q1 and no OP 19.6 28.9 39.6 65.6 25.6 36.1 52.5Q2 and OP 27.0 37.8 49.2 74.1 34.0 45.5 62.1Q2 and no OP 19.6 28.9 39.6 65.6 25.6 36.1 52.5Q3 (with or without OP) 25.9 36.5 47.8 73.0 32.7 44.2 60.8Q4 (with or without OP) 30.6 41.8 53.4 77.4 37.9 49.7 66.0
Note:OP = member of an occupational pension schemeQ1 = top quartile of the wage distribution
Older age groups are more likely to be displaced (within the six-year period) than younger ones. Women are more likely to bedisplaced than men of the same age.7 The following results hold, after taking age and gender effects into account.
• People in the bottom quartile of the hourly wage distribution are at the greatest risk of displacement;
• People in the top half of the wage distribution have a reduced chance of displacement, provided they do not have anoccupational pension;
• Being in a job with an occupational pension significantly increases your chances of being displaced if you are in the tophalf of the wage distribution, often by ten percentage points or more. Men in their early 50s with an occupational pensionand in the top quartile of the wage distribution are 50% more likely to be displaced than a man with the same age andhourly wages but no occupational pension. The effect on the third quartile is very weak and on the bottom quartile non-existent;
• These results hold even after controlling for the fact that some people initially worked in industries that faced sharpemployment falls during the period, while other industries grew. Displacement was, unsurprisingly, higher among peopleworking in an industry whose employment fell by more than 2 per cent a year on average between 1990 and 1995;
• Although working in a shrinking industry increases the risk of displacement, working in a growing industry does not reducethe risk compared to working in one whose employment is static;
• Finally, sector appears not to have a significant additional effect. The private sector, the civil service, local authorities andthe NHS all seem – after controlling for age, gender, wages, occupational pensions, and changes in industry employment –to perform similarly with respect to the displacement of older workers.
56
6 Further details of probit regressions are in Campbell (1999)7 This is probably related to the fact that the state pension age for women is lower than that for men.
Few people return to the labour market
One reason for the lower employment rates is that relatively few older people return to work after being out of the labourmarket. About one-quarter of the panel of people over 45 returned to work after being displaced. The return rates are evenlower (18%) for people who were not employed at the start of the survey period. And nearly half of those 18% are back out ofwork before the end of the period. Table 3 has the details.
Table 3: Proportions returning to work (%)Men Women Total
Age at beginning: 45-49 50-54 55-59 60-64 45-49 50-54 55-59
% of those not initially employed, who during the sample period:return to employment at some point 22 26 11 9 35 19 11 18return and stay in work 17 17 6 4 24 10 7 11
return but don’t stay 6 9 5 5 11 10 4 7never work in period 78 74 89 91 65 81 89 82
% of displaced who:return to employment 49.0 35.1 10.7 10.9 31.6 22.8 12.6 24.1don’t return 51.0 64.9 89.3 89.1 68.4 77.2 87.4 75.9
Memo items:% not employed initially 9.5 18.1 30.2 47.7 27.2 32.6 46.9 29.3% employed initially and are later displaced 25.9 31.5 38.4 41.4 22.9 32.0 33.1 31.2
Tables 4 and 5 show that there were significant movements within the broad category of being without work, and suchmovements were generally associated with lessening labour market attachment. About half of those who were unemployed atthe start of the survey period were either long-term sick or retired (or, for women, looking after their family) by the end. Aboutone-third of the long-term sick had retired by the end.
Of those who did return to work, unemployed people were more likely to return to work than people who were long-term sickor retired. The long-term sick had even less chance of returning to work than those who were retired.
Table 4: Men’s transitions between different labour force states (%)
Labour force stateinitially: Labour force state at end of BHPS period
Employed Unemployed Long-term sick Retired Student Totalor other
Employed 67 5 6 22 0 100Unemployed 19 26 23 31 1 100Long-term sick 3 3 62 31 1 100Retired 5 0 2 93 0 100
Table 5: Women’s transitions between different labour force states (%)
Labour force stateInitially: Labour force state at end of BHPS period
Employed Unemployed Long-term sick Retired Looking after family Total
Employed 66 4 4 20 6 100Unemployed 33 15 4 19 30 100Long-term sick 4 0 58 31 6 100Retired 6 3 1 74 16 100Looking afterfamily 15 1 6 24 54 100
57
What do the results tell us about the explanations of falling employment among older workers?
There are five explanations for the increase in employment among older workers:
• More people making a voluntary choice to retire early;
• Labour supply reductions that are involuntary or the result of either constrained choices or distorted incentives;
• The effects of occupational pensions;
• A shift in labour demand away from older men;
• Increasing age discrimination.
It is sometimes difficult, even in individual cases, to differentiate between these explanations. However, we can draw someconclusions on the basis of the evidence presented in the paper.
We can infer that age discrimination is not likely to have been the major cause of the dramatic fall in employment among olderworkers. Discrimination – in the sense of unequal treatment of people who could do the job equally well – on the grounds ofage seems to have been experienced by relatively few older people. McKay and Middleton (1998) found that 5 per cent ofpeople aged 45-69 believed that they had been discriminated against in one or more job applications because they were felt tobe too old. However, the arguments against age discrimination hold even if, as the evidence suggests, it is not widespread.
Occupational pensions are associated with lower employment rates. People whose wages are in the top half of the distribution,are more likely to leave the labour market if they have access to occupational pensions. Occupational pensions could bereducing employment in a number of different ways. 90% of occupational pensions are salary-related, and employers’contributions to such pensions increase, often substantially, as people near retirement age. They therefore add to the cost ofemploying older workers and provide incentives on employers to encourage their employees to retire early. Separately,occupational pensions may reduce labour supply, either voluntarily or as a result of constrained choice. The first case seemsbenign, with greater pension entitlements making people richer and enabling them to choose to retire earlier. Sometimes,however, early retirement packages, including an enhanced occupational pension, are offered on a temporary basis when thefirm is seeking to reduce its workforce. People accepting that offer may be making what is a very constrained choice,especially if they feel that redundancy later is the likely alternative.
There is also evidence of a labour demand shift against older men, as this group has seen the largest decline in theiremployment rates and a fall in their relative wages over the last twenty years. This shift has taken place over a long period, andso does not reflect the level of aggregate demand at a particular point in the economic cycle. Older men were more likely tohave been working in industries whose employment declined dramatically in the early 1990s. Men over 45 were much morelikely to have been working in shrinking industries. Half of them worked in industries whose employment fell by 12% or morebetween 1990 and 1995, while aggregate employment fell by less than 4%.
That leaves labour supply explanations for the changes. To what extent did labour supply reduce voluntarily? And to whatextent were labour supply reductions involuntary or the result of constrained choices?
As successive generations are richer than their predecessors, theory would suggest that voluntary early retirement mightincrease. This explanation is likely to hold for a proportion of early retirees, although not all the increase in displacementassociated with occupational pension schemes will be voluntary.
However, voluntary, unconstrained decisions are, at best, probably a limited description of the story. There are a number ofreasons to believe that much of the change in employment reflects constrained choices or involuntary decisions:
• higher levels of wages are associated, in the absence of occupational pensions, with lower displacement rates. Voluntary,unconstrained decisions are not likely to have caused the substantial falls in employment among men at the bottom of thewage distribution;
• older men were more likely to be working in industries which shrunk in the early 1990s. Accepting an offer of earlyretirement in such an industry may be a constrained choice;
• a number of older people move from unemployment to long-term sickness or to retirement, and from long-term sickness toretirement. People making a voluntary decision are likely to move straight from employment to retirement instead;
• people who have been displaced typically find that their potential earnings in a new job are significantly less below theirprevious wages. This “pay gap” is usually larger for older workers. People who do not re-enter the labour market becauseof this pay gap are not making an unconstrained choice;
• the scale of the problem – with the proportion of older men out of work increasing from one-fifth to two-fifths – is too largeto have been caused solely by a number of individuals deciding voluntarily to retire early.
58
What about the future?
Older men have been particularly affected by the fall in employment. There are two types of problem here. The obvious oneaffects people currently in their fifties: older people unexpectedly without work are likely to see both their current income andtheir income in retirement affected. The other issue is about what will happen in the future when younger people in the labourmarket reach their fifties.
The evidence presented here, covering 1979 to 1997, do not give grounds for confidence that the trend to lower employmentamong older men will reverse of its own accord (although employment might not be expected to fall as quickly as between 1979and 1983). Indeed, later cohorts have lower employment rates, even at age 40, than men born before them. This implies that menin their fifties in the future could be even less likely to be employed than the current generation of older workers.
Older people without work are more likely to stay that way than younger people. Return rates to work are lower for theeconomically inactive than the unemployed, and for older people than the young. The evidence is that the labour marketattachment of a workless older person tends to weaken over time, reducing further the likelihood of their later returning to ajob. We have also seen that falls in male employment have been accompanied by an increase in the number of people who areeconomically inactive, and are therefore not looking for or available for work.
The main findings of this paper have been that:
• employment has fallen sharply among men, especially men over the age of 50;
• economic inactivity, much more than unemployment, has increased at the same time;
• the trend seems to be continuing, with each successive generation more disadvantaged than its predecessors;
• economic recovery will not solve the problem on its own, according to regional data;
• two groups of older workers have been most likely to leave the labour market. One group consists of people in the bottomquartile of the wage distribution. The other is people in the top half but who are members of an occupational pensionscheme;
• few older people return to work after leaving the labour market. Indeed, older people without a job are more likely tobecome less attached to the labour market over time, moving from unemployment to long-term sickness or retirement orfrom long-term sickness to retirement.
Older women have not shared in the increase in female employment. This may be due to a cohort effect and if so, that might beless of a long-term concern. Looking to the future, the remaining concerns might then arise from:
• lower male employment,
• the sharper falls among older men, implying a trend of more people leaving the labour force early,
• the possibility that the next generation of men will stop work even earlier than the current generation, and
• whether any of the causes of lower male employment will limit the effective retirement age of women in the future.
ReferencesCampbell, N (1999). ‘The Decline of Employment Among Older People in Britain’ CASEpaper 19. London School of Economics: Centre for Analysis ofSocial Exclusion.McKay, S and S Middleton (1998). ‘Characteristics of Older Workers’, DfEE Research Report No. 45. London: The Stationery Office.Paull, G (1996) ‘The Biases Introduced by Recall and Panel Attrition on Labour Market Behaviour Reported in the British Household Panel Survey’, Centrefor Economic Performance Working Paper no. 827. London School of Economics.
59
60
‘Prioritising older workers’
Richard Disney, Professor of Economics, University of Nottingham, and Research Fellow, Institute for Fiscal Studies.1
Introduction
A serious problem for many European countries, both in OECD and in Eastern and Central Europe, is the growing rate ofeconomic inactivity among older people, especially older men. In the UK, for example, there are over 2,500,000 people agedbetween 50 and state pensionable age (currently 65 for men and 60 for women) who are economically inactive. And inaddition, among those who define themselves as ‘economically active’, many are in fact never likely to work again, havingbeen unemployed for many years.
Many inactive people have retired ‘voluntarily’ through an occupational pension scheme. Early retirees are heterogeneous, andit should not simply be assumed that early retirement from paid work is automatically associated with poverty in old age, forexample. Early exits from the labour market may be associated with both positive and negative income shocks. An exampleof the latter is inactivity on grounds of ill-health.
An opposite error is to assume that early retirement, even with an occupational pension, guarantees an adequate income. Manyprospective occupational pension entitlements are rather small (Disney and Whitehouse, 1996), and employers in final salaryoccupational pension schemes have every incentive to create early retirement ‘windows’ to eliminate high cost staff sincepension and pay profiles are ‘backloaded’ in final salary schemes. And for those without occupational pensions, incomeadequacy may be a severe problem. Not surprisingly, the 1980s and 1990s, which have seen a retrenchment in public pensionprovision, a growth in private pension income receipt, and two severe recessions, have also seen significantly wideningpensioner inequality (Johnson and Stears, 1995).
Work by Tanner (1998) and by Disney and Tanner (1998) reinforce the view that the dynamics of retirement are complex andthat individuals retire ‘prematurely’ both because of positive and negative shocks. For example, Disney and Tanner (1998) useeconometric techniques to examine probabilities that individuals subsequently retire earlier, later or at the time that theyexpected, using the two waves of the Retirement Survey. Being unemployed and becoming disabled are strongly associatedwith retirement earlier than expected; lower income and wealth weakly so. These are all cases where there may be a linkbetween retirement behaviour and subsequent poverty. But having an occupational pension increases the hazard of labourmarket exit after age 55, and lacking educational qualifications makes it less likely that individuals will retire early thanexpected ceteris paribus. Likewise, women who become divorced or widowed are likely to retire later than expected: theseare examples of where negative income shocks induce individuals to remain in the labour market for a longer period thanoriginally anticipated.
Some evidence
Charts 1 and 2, taken from Labour Market Trends, show the participation rates for the UK for men and women aged 55-64from 1976 to 1993, projected to 2011. For men, there has been a steady decline in the rate of labour force participation forthose aged 60-64 and around 50% of that male age group are now in the labour force. For those aged 55-59, there was anespecially sharp fall in the early 1980s. Statisticians project a continued downward trend, but at a less precipitate rate.
For women, there is a different picture, with roughly stationary participation in the past, and even a projection of an increase inparticipation for the 60-64 year olds in the early part of the next century (which is in part linked to the raising of statepensionable age for women from 60 to 65 between 2010 and 2020). However the trend for women conceals two conflictingfactors: a rise in cohort-specific participation rates over time, accompanied with an earlier downturn in the activity rate by agefor each cohort, albeit from a higher base (see Disney, 1996, Chapter 7).
It should also be reiterated that economic ‘activity’ is not the same as employment, and many older people (especially men) areunemployed or on benefits prior to switching to incapacity benefits. Indeed employmentrates of men and women in the 55-59
1 My thanks to seminar participants at the Centre for Economic Performance, LSE, and at the World Bank, for my comments on earlier drafts of thispresentation.
Labour Force Participation RatesMen Aged 55-641974 1994
Australia 80 60Canada 80 60France 70 42Germany 73 52Italy 42 30Japan 86 86
Labour Force Participation RatesMen Aged 55-641974 1994
Holland 75 43Norway 82 72Portugal 82 65Spain 81 57Sweden 82 76US 78 66
Table 1 Male Labour Force Participation in various OECD Countries 1974-94
61
age category are fast approaching equivalence, at around 50% of the age group. Finally, Table 1 shows a similar process goingon in other OECD countries although there are interesting outliers, such as Japan. These data are taken from OECD (1996),where a number of explanations are adduced for the differences between countries; see also Gruber and Wise (1997).
Explanations of the trend
The decline in participation can be linked naturally to two sources of ‘shocks’; from the demand side or from the supply side.If the source is demand shocks, it is necessary to ask why older workers (especially men) have been hit disproportionatelyhard. If the source is supply shocks, the question is whether certain policies do in fact impinge on this age group. For example,the standard policy advocated in OECD (1996) and by Chand and Jaeger (1996), of raising state pensionable age as a means of
62
improving the support ratio of workers to pensioners, may be quite irrelevant to participation if a large fraction of olderworkers have already effectively retired from the labour market long before reaching state pensionable age.
Figure 1 provides a stylised account of why demand shocks hit older workers. Here it is assumed that the age-specific marginalproduct of labour, f ’(l) , is quadratic, reaching a maximum in late middle age. The combined wage plus accrual of SocialSecurity Wealth (SSW) is positively related to age. Thus a downward shock to the demand for labour, reflected in a shift of themarginal product from 1 to 2, reduces the retirement age. Note (Figure 2) that where there is a backloaded pension plan, so thatthe age-remuneration profile is non-linear, ‘normal’ retirement will take place at an earlier age but the volatility of retirementage in response to demand shocks will be lower.
In Figures 1 and 2, it is the combination of the age-specific marginal product of labour and the remuneration profile that togethergenerate changes in the retirement age in the face of demand shocks. However there are additional strong reasons for thinkingthat older workers will be disproportionately affected by changes in labour demand. First, there are more old workers relative toyoung workers at present as a result of the ‘baby boom’ generation reaching middle age. Second, insofar as the source of demandshocks has been technological, reflected in skill-biased technical change, older workers lose out if the skill shift is specific tocohorts entering the labour market, who are more willing and able to acquire new skills. Older workers are disproportionately hitby technical change biased towards the skills of labour market entrants, both because older unskilled workers are close substitutesfor young unskilled workers and because older skilled workers are not close substitutes for young skilled workers.2
The case of a downward shock to labour supply, identified here as a shift in the marginal utility of leisure, u’(leisure),from 1 to2, is depicted in Figure 3. Note that where the remuneration profile is backloaded (as in the dotted curve), it is hard to believethat a supply explanation is plausible, because the marginal return to staying in work may outweigh the marginal disutility ofwork. This is of course the basis of the famous Lazear (1979) argument concerning mandatory retirement.
Although Figure 3 depicts the supply factor as a rising ‘taste’ for leisure, there is an influential literature, represented by Gruber
2 For a discussion of production functions which attempt to separate age and education substitutability and complementarity, see Disney (1996), Chapter 6.
f1(1)1
f1(1)2
f1(1)1
f1(1)2
63
and Wise (1997) which argues that it is the impact of incentives on the after-tax age-remuneration profile that induces earlyretirement. Such factors include actuarially favourable early retirement, which induces ‘spikes’ in the profile (for anillustration, see Kotlikoff and Wise, 1985) and other taxes on subsequent work. A stylised version of a profile which generatesearly retirement ‘spikes’ is given in Figure 4, which shows distinct ‘spikes’ at 60 and 65. However it should be noted that, inBritain, many of the factors adduced for such non-linearities elsewhere, including explicit early retirement provisions in thepublic pension programme and explicit taxes on pensions if individuals continue to work, do not exist.3 However, suchperverse incentives have existed in the past in the UK, such as when it was possible to retire on tax-free Invalidity Benefit,which was therefore more attractive than retiring with the Basic State Pension at 60/65.
Policies
It remains to consider policies that might reverse this trend towards early departure from the labour market. On the labourdemandside, there are some positive and negative conclusions:
• do not use early retirement as a means of cutting labour supply, whether by overgenerous retirement provisions in the socialinsurance system or by a disability programme. This strategy builds up future trouble for the viability of the social securityprogramme and is an expensive alternative to more active policies. For example, letting an individual man leave the labourmarket at age 50, with little chance of re-entry at a later stage, will have a present value cost of well over £100,000 inbenefit payments and lost tax receipts;
• work sharing and limits on the working week do not overcome the skill bias problem;
• toughening-up age discrimination legislation is problematic, even though ‘codes of conduct’ and educational programmes
3 For further details, see OECD (1996) and Gruber and Wise (1997).
for employers are essential. Ceilings on ages in job advertisments are already illegal as a form of indirect discrimination.But younger and older workers do have different skills and attempts to compare equivalents will produce a large income foremployment lawyers. There is evidence from the United States that older workers are hired into a smaller sub-set ofoccupations and industries due to skill-specificity, the costs of final salary pension plans, and so on. These may be themargins on which to operate;
• training and retraining therefore seem more attractive policies. Simply paying employers to retain older workers withouttraining has high deadweight costs (as in the temporary employment subsidy programmes of the 1970s) and public trainingof out-of-work men has a low return. Subsidies to firms which are linked to explicit training and retraining programmes forolder workers would seem to be the answer. This may need a refocus of New Deal policy: although the private returns totraining young people are much higher, this differential provides no rationale for public intervention, which should arisewhere there is a wedge between private and social rates of return;
• re-examine the tax treatment of final salary pension schemes which use premature retirement as a means of downsizing theworkforce: in the short term it is attractive for pension funds to be so used but in the long run the burden of such policieswill be borne by other scheme members and by the state. Explicit use of such schemes may ultimately lead to youngerworkers ‘voting with their feet’ and initiating other pension arrangements, such as Personal Pensions.4
On thelabour supplyside:
• encourage transparent pension arrangements. Many people do not appear to understand the conequences of annuitisationand in particular of incomplete indexation. Provision of large tax-free lump sums and a high initial pension value mayinduce people to overestimate the value of their future pension stream;
• avoid all non-linearities in the benefit system which encourage premature retirement: actuarially favourable earlyretirement, tax-free treatment of certain benefits, excessive taxes on pensions if individuals continue to work after statepensionable age, and so on. The fact that all these policies have been adopted in the UK since the mid-1980s does howeversuggest that the problem of premature economic inactivity cannot simply be attributed to the ‘wrong’ incentives –maintaining the demand for labour is also important.
ReferencesChand, S. and Jaeger, A. (1996) Aging Populations and Pension Schemes,Occasional Paper #147, International Monetary Fund, Washington.Disney, R. (1996) Can we afford to grow older? A perspective on the economics of aging,Cambridge: Mass, MIT Press.Disney, R. and Tanner, S. (1998) ‘Retirement and retirement expectations in Britain’, mimeo, Institute for Fiscal Studies.Disney, R. and Whitehouse, E. (1996) ‘What are occupational pension plan entitlements worth in Britain?’Economica,63, may, 213-238.Gruber, J. and Wise, D. (1997) ‘Social security and retirement around the world’, National Bureau of Economic Research, Working Paper#6134, Cambridge:Mass.Johnson, P. and Stears, G. (1995) ‘Pensioner income inequality’, Fiscal Studies,16, November, 69-94.Kotlikoff, L. and Wise, D. (1985) ‘Labour compensation and the structure of private pension plans’, 55-88 in D. Wise (ed) Pensions, Labor and IndividualChoice,Chicago UP for National Bureau of Economic Research.Lazear, E. (1979) ‘Why is there mandatory retirement?’Journal of Political Economy,87, 6, 1261-1284.OECD (1996) Ageing in OECD Countries: A Critical Policy Challenge,Social Policy Studies, No. 20, Paris.Tanner, S. (1998) ‘The dynamics of male retirement behaviour’Fiscal Studies,9, May, 175-196.
64
4 For a discussion of the viability of final salary pension plans when the choice of defined contribution instruments is available, and where the workforce isageing, see Disney (1996), Chapter 5.
65
‘UK Pensioner Incomes’
Carl Emmerson, Institute for Fiscal Studies
Introduction
When looking at income dynamics and the importance of dynamic income inequality, pensioners are a group that are ofparticular importance. Not only because once individuals have reached retirement they are extremely likely to remain on thesame income level, but more importantly because for many pensioners that level of income leaves them low down in theincome distribution.
This short paper starts by describing the forms of income received by pensioners, both from the state and privately. It then goeson to look at the level of income in retirement this is presently providing, where it leaves pensioners in the overall incomedistribution and also the level of income inequality between pensioners. It then concludes with the implications of the presentsystem for future generations of pensioners.
Current pension provision
Pensioners in the United Kingdom currently receive significant proportions of their income from both the state and privatesources of income. The extent of the role that the state plays in the provision of retirement income influences both the level ofprovision made privately, and the level of income inequality amongst the elderly. This section describes the current system ofstate provision for retirement incomes and how, without further reform, this will change in the future before going on todescribe the private forms of income both received by pensioners today and that which is expected to be received by futuregenerations of pensioners.
Public pension arrangements
State provision is essentially in two tiers. The first consists of the Basic State Pension, and the increasing importance of means testedbenefits. The second tier is the State Earnings Related Pension Scheme which is more commonly known by its acronym SERPS.
Basic State Pension
The basic state pension is a flat rate contributory benefit. It is easily the most expensive single item of government spending,currently costing £32 billion p.a. (3.7% of GDP), which is about one third of total social security spending. It is currently worth£64.70 p.w. for a single pensioner, which is equivalent to just 15% of average male earnings. Figure 2.1 shows how the valueof the basic state pension in real terms and as a percentage of average earnings has changed since it was introduced in 1948.Between 1948 and 1982 it often increased in value by more then the growth in earnings, before it was formally indexed toprices in 1982. Real earnings growth since 1981 has led to its value as a share of average earnings falling from 20% to around15% today. If price indexation continues then the basic state pension will be worth under 7% of average earnings by the middleof the next century.
Receipt of the basic state pension is conditional on contributions being credited, and whilst the vast majority of men aged overthe state retirement age qualify this is not true of women. In future however receipt will become nearly universal as morewomen will qualify on the basis of their own contribution record. This is due to three factors:• Increased labour market participation by women;• Contribution credits are now given for time spent being in education or training, or for periods of unemployment or
disability;• The introduction of Home Responsibility Protection in 1978 which means that the number of years for which contributions
are required is reduced for years spent looking after children or sick relatives.
Source: HM Treasury (1996)
Figure 1: The level of the Basic State Pension in relation to earnings and real prices (1948 to 1995).
1948 1953 1958 1963 1968 1973 1978 1983 1988 1993
66
Means tested benefits
A significant proportion of pensioner income now comes from means tested benefits. This consists mainly of Income Supportwhich is the principal benefit for those with low incomes; Housing Benefit which is a payment towards the housing costs ofthose in rented accommodation and Council Tax Benefit which provides assistance towards the cost of local taxes. The total costof government spending on means tested benefits to those aged over the state retirement age is £9.6 billion (1.2% of GDP).
Currently one in four Income Support claimants is aged over the state retirement age, with the majority of these being singlewomen. In total around 15% of those aged over the state retirement age qualify for Income Support. The principal reason forthis is that the Income Support level is now, and has been for many years, at a higher level then the basic state pension. This isan effect of government policy to only uprate the basic state pension in line with prices, whilst Income Support has risenslightly faster. The recent announcements of a Minimum Pension Guarantee following the Government’s ComprehensiveSpending Review indicate a continuation of this policy. Whilst this undoubtedly ensures that additional funds are concentratedon the poorest pensioners, it also means that those approaching retirement with low levels of saving have little incentive tomake any provision for themselves, since Income Support is withdrawn with a 100% taper.
State Earnings Related Pension Scheme (SERPS)
Although it was only introduced in 1978, perhaps the most notable feature of SERPS is how short lived its original form was.It was introduced to provide compulsory second tier pension coverage for all those earning above the Lower Earnings Limit(LEL) who did not have an occupational pension scheme. Whilst under its original format it was quite generous, subsequentreforms have substantially reduced the amount of entitlement and hence the size of future SERPS payments.
Private pension arrangements
There are two main types of private pension schemes – occupational pension schemes and personal pension schemes. The first,which are by far the most important for the current generation of pensioners, typically operate on a defined benefit basis. Forexample a typical scheme may provide a pension equivalent to 1/60th of final salary for each year worked. Around 50% of theworkforce are members of such a scheme, with higher levels of coverage amongst older workers. This has been relativelyconstant for around 30 years, so it is not the case that in future more individuals will retire with occupational pensions.However entitlements will continue to rise in future, despite the lack of increase in coverage, since younger generations will,due to real earnings growth, have higher lifetime earnings.
Personal pension schemes, although not being a particularly important source of income for the vast majority of today’spensioners, will become more important in future. They typically operate on a defined contribution basis, and were allowed asan alternative to SERPS from 1988 onwards. They are particularly popular amongst the young, who have longer to accumulatereturns in these funds. In addition the tax system encourages saving for retirement in this form.
Pensioner incomes
So given the system of state and private pension provision at present in the United Kingdom, what does this actually providefor the current generation of pensioners? This section looks at the incomes of different groups of pensioners, how thiscompares to that of the population as a whole, and the distribution of income between pensioners. This is done using data fromthe 1995 Family Resources Survey, which is an annual survey of around 27,000 households containing approximately 47,000individuals. For the purposes of this work a pensioner has been defined as someone aged over the state retirement age, orsomeone aged over 60 & not employed or self employed. Couples have been defined according to the status of the husband.The measure of income used is the total net income to the benefit unit.
Table 1 shows how average incomes of pensioners vary between different types of pensioners and different age groups.Pensioner couples and single male pensioners tend to have higher average incomes once they reach 65, which is due to mennot being able to claim the basic state pension until they reach that age. With this exception older groups have, on average,lower incomes. This is unsurprising since as mentioned earlier younger pensioners will have higher levels of a averageearnings due to real earnings growth.
Table 1: Average U.K. pensioner income (£ per week), by status and age group.
Age group Status 60 – 64 65 – 69 70 – 74 Over 75 All over 60Couples 228 254 219 206 228Single men 125 150 132 127 133Single women 139 126 113 109 116
Source:1995 Family Resources Survey; Emmerson and Johnson (1998).
When assessing the economic well being of pensioners, it is the comparison of their incomes to that of the working populationthat is likely to be of greater importance. Figure 2 shows what percentage of pensioners are in each income decile, so if
67
pensioners were equally spread across the income distribution there would be 10% in each one. What we actually find is thatthere are under 10% of pensioners in the first income decile, which since the start of the 1970s has increasingly being taken upby growing numbers of unemployed (Goodman and Webb, 1994). However pensioners are significantly over represented indeciles two through to five, but not uniformly poor, since there are still significant proportions in the top four income deciles.This is a different picture to that of the 1970s with pensioners much more concentrated in the lower income deciles (Johnsonand Stears, 1995).
Figure 2: Proportion of pensioners in each income decile.
Note:Benefit unit income equivalised to that of a single person using the (1,0.7,0.5) equivalence scale. For example the income of a couple with no childrenwill be divided by 1.7, whilst that of a couple with two children will be divided by 2.7.
From a dynamic perspective pensioners are clearly a different group to the rest of the population, since once retirement hasbeen reached they are likely to remain on the same level of income for potentially a substantial period of time. This means thatdespite evidence that it is tending to be other unwaged groups such as lone parents and the unemployed that are in the firstincome decile these individuals will be more likely to be in higher deciles at a later date than pensioners. Hence the presence oflarge proportions of pensioners in the lower income deciles may be more concerning then for other groups. Of course it couldbe the case that due to dissaving consumption by the retired was actually much higher then their current incomes. However, asdiscussed in Banks et al (1998), upon retirement consumption actually tends to fall faster then income.
Figure 3 looks at the income distribution of pensioners, and how the composition of that incomes varies. This gives a picture of what is behind the much wider dispersion of pensioner incomes compared to the 1970s. Looking first at the levels of average income between the five quintiles there is only a small difference between the first four quintiles,but with substantially higher levels of income in the fifth quintile. Average receipt of the basic state pension is roughly constantacross the quintiles, which is unsurprising given its flat rate nature. Perhaps more surprising is the significant receipt of means-tested benefits amongst those in the top quintiles. This reflects the fact that for some people means-tested benefits, especially Housing benefit, are generous enough to lift them up the distribution. The main factor behind pensionerincome inequality is the much higher levels of income from private pensions, investments and also earnings in thetop two, but particularly the richest, income quintiles. Average private pension income by those in the richest income quintile is actually more then total average income for those in the second income quintile. In future, given the reductions inthe generosity of SERPS and the fact that it will be those with higher life time earnings who are able to save significant amounts in personal pension schemes, the trend of increasing income inequality amongst pensioners is likely tocontinue.
Figure 3: Equivalised net pensioner income, by income quintile.
Note: The individual components have been calculated by multiplying the percentage of total income that they correspond to with the average net equivalisedincome for each quintile. Source: 1995 Family Resources Survey; Emmerson & Johnson (1998).
Policy conclusions
Reforms over the last twenty years, coupled with a less stark demographic situation have left the UK in a position, whereunlike other European neighbours the state pension system can be afforded without future increases in taxes. However, whilstthe present system is sustainable in terms of its affordability, it remains to be seen whether it is sustainable in terms of itsacceptability. Whilst increasing the level of Income Support to pensioners rather than the Basic State Pension does redistributeto the poorest pensioners it also provides little incentive for those approaching retirement with little savings to try and makesome provision for income in retirement.Without further reforms, the future is likely to see two groups of pensioners emerging. Those who have been able to makeenough savings for retirement to have significant levels of income from private sources, and those who are reliant on statesources of income. Whilst more compulsion would lead to increased savings for some, for many the reason individuals areunable to save significant amounts will be that they have insufficient earnings during their working lives. It is unlikely that anyreasonable degree of compulsion will stop these groups having low levels of income in retirement.
ReferencesBanks, J., Blundell, R. and Tanner, S. (1998), ‘Is There a Retirement-Savings Puzzle?’, American Economic Review,vol. 88, no. 4, pp. 769-788.HM Treasury (1996), Tax Benefit Reference Manual, 1995-96 edition,London: HM Treasury.Department of Social Security (1998), Social Security Departmental Report: The Governments Expenditure Plans 1998-99.London: The Stationary Office.Dilnot, A., Disney, R., Johnson, P. and Whitehouse, E. (1994), Pensions Policy in the UK: An Economic Analysis,London: Institute for Fiscal Studies.Emmerson, C. and Johnson, P. (1998), Pension provision in the United Kingdom,Institute for Fiscal Studies (mimeo).Goodman, A. and Webb, S. (1994), For richer, for poorer: The Changing Distribution of Income in the United Kingdom, 1961-91,Commentary no. 42,London: Institute for Fiscal Studies.Johnson, P. and Stears, G. (1995), “Pensioner Income Inequality”, Fiscal Studies,vol. 16, no. 4, pp. 69-93.
68
69
Session 3: Retirement and the Elderly: Discussion
The earlier sessions of the conference had stressed the need for early intervention to prevent poverty but this was clearly moredifficult in the case of poverty in retirement. Today’s pensioners began their working career before the pension system whichwe have in place today existed. By the nature of the problem, poverty amongst the retired would require alleviation for today’spensioners as well as prevention for future pensioners.
Over the 1980s there was a tendency to disregard the decline in employment rates of older men as part of a decline inmanufacturing and a general increase in unemployment. The presentations had clearly shown that there was a secularcomponent which had changed the labour market opportunities of older male workers.
These problems raised some policy issues;
• How does the state influence the individual’s decision to retire? In some cases early retirement is through choice rather thandisplacement, how does the state value leisure?
• How can the different trends in employment rates of men and women be explained? There may be a certain logic to thesetrends where couples wish to retire together.
• How should the Government intervene? The German model where there are moves towards encouraging early retirementruns counter to UK policy.
• Should the Government be prescribing one occupational pension over another? Is there a trade-off between final salary andmoney purchase schemes, whereby the former might decrease employment rates among older workers, but the latter mightleave pensioners worse off financially?
• Does the link between final salary related schemes and early retirement imply a need for “winding down”?
• How does the decline in employment rates of older men tie in with lifelong learning? Should we continue to train our olderworkers?
• Should the Government be providing in-work subsidies to older workers, along the lines of the Working Families TaxCredit?
• The decline in employment rates has been equally matched by an increase in inactivity rates for older men, this has meant anescalation in the number of incapacity benefit claimants. Have the reforms to the all-work test and incapacity benefit receiptbeen enough/too much?
• Does the link between poverty and longevity mean that poverty in retirement is being understated?
• How can the Government strike the right balance between providing minimum income guarantees and creating incentives tosave for ones own retirement?
During an open discussion the following issues were raised;
Demand for older workers in the labour market
Do older workers lose jobs at a faster rate than younger workers? It is clear that older workers are less likely to return to workthan younger workers following unemployment or inactivity. It is more difficult for older workers to find jobs because theytypically have longer durations and because they have to face a more concentrated search for employment. Someoccupations/employers will not take on older workers, either because of backloaded defined benefit pensions schemes orbecause of the costs of re-training older workers.
Replacement ratios and the distribution of income
Measured or stylised replacement ratios can exceed 100% by a considerable amount. But the greatest income gains to retiringare not always found for those on occupational pensions. The greatest income gains are typically to married women who are onlow incomes before retirement. Replacement ratios are complicated across the income distribution. Moreover the distributionof occupational pensions is very wide, there are many people with only very small occupational pensions. A pension schemewhich looks good on paper can provide quite different returns if earnings and contributions are only intermittent. Therefore,having an occupational pension does not necessarily equate to remaining out of poverty in retirement. Furthermore, providinga high occupational pension means that there is every incentive for a firm to unload its older workers.
What about distribution across the family - if the employment rates of older women are increasing and more women have apension in their own right then should we worry about the decline in male participation? Some of the trends in older workers’employment rates is attributable to personal choice but this does not mean that there is not a problem. There is clearly a demandside problem whereby it is harder for older workers to get back into the labour market even if they want to continue working.
Influencing the decision to retire
At present the government effectively subsidises the decision to retire relative to the decision to continue working becauseolder workers continue to pay tax whilst the retired do not. Whilst there might not be a sensible way in which to equalise thischoice, one possibility which was suggested was to raise awareness of the issue. For example, bringing attention to the subsidyinvolved in mortgage interest relief helped to focus the policy debate.
69
Retirement age
Subsidising early retirement, as in the German model, would run counter to UK policy on retirement. There is a potential rolefor the public sector to play in the debate about early retirement. Public sector workers are some of the worst offenders forretiring early, the proportion of workers retiring on ill health grounds is 39% in local government, 20% in the civil service and25% of teachers whilst the median private sector firm was British American Tobacco with 9% of their employees retiring onill-health grounds. Early retirement in the public sector was seen as particularly problematic given that most public sectorpension schemes are unfunded.
Training older workers
Should the Government be subsidising retraining of older workers? Employer based training appears to be better thangovernment training in this context, the internal rates of return are highest to in-house training provided by the employer. Thedeadweight costs of subsidising retraining of older workers must be high but it is likely that the pay-off would be better than asubsidy purely for retaining the older worker. The private return to training younger workers is inevitably higher than totraining older workers but perhaps the policy prescription about who to train should not be dependent on private rates of return.The fact that there seems to be a general trend for frequent retraining must help older workers since the incentives to train olderworkers increase as the returns increase.
70
Session 4: Work and Poverty
‘Low pay, no pay dynamics’
Mark B. Stewart, University of Warwick
Introduction
Dynamics are important when investigating low pay. Some of the individuals who are low paid in one period are paid abovethe threshold the following year. Others are no longer employees. Similarly those not working and those above the paythreshold in the previous period may move into low pay. This paper looks at these transitions into and out of low pay togetherwith movements into and out of employment, i.e. at the dynamics of low pay and no pay.
Low pay persistence
Starting with the customary approach when examining earnings mobility, transitions within the earnings distribution areexamined, restricting attention to those who are employees at the start and end of any transition considered.1 Estimatedconditional probabilities are presented in Table 1. The degree of persistence exhibited is strong. The probability of being lowpaid in year t is much higher if the individual was low paid at t-1. For the £3.50 threshold the probability of being low paid at tgiven low paid at t-1 is 57% compared with 3% given higher paid at t-1 They are similarly far apart for the other twothresholds.
Table 1: Conditional Low Pay Transition ProbabilitiesThreshold
£3.50 £4.00 £4.50
Probability of being low paid at t given:Low paid at t-1 .57 .68 .74Higher paid at t-1 .03 .05 .06
Probability of being low paid at t given low paid at:t-1, t-2 .73 .79 .83t-1, t-2, t-3 .72 .84 .85t-1, t-2, t-3, t-4 .77 .83 .85
Probability of being low paid at t given low paid at t-j-1 but higher paid at each of the j years in between:j=1 .27 .28 .30j=2 .10 .15 .19j=3 .10 .11 .12
Probability of being low paid at t given higher paid at:t-1, t-2 .02 .03 .04t-1, t-2, t-3 .02 .02 .03t-1, t-2, t-3, t-4 .01 .02 .02
Source: Stewart (1998a)Data: BHPS, Waves 1-5
Earnings adjusted to April 1997
Table 1 also indicates that the degree of persistence is higher for those who have already been low paid for more than oneperiod. Looking first at conditioning on two periods, the probability of remaining low paid is higher if the individual was lowpaid at t-1 andt-2 than just conditional on t-1: 73% against 57% in the case of the £3.50 threshold for example. Additionalconditioning on being low paid at t-3 and t-4 results in a much smaller change in the probability of remaining low paid.
For those above the low pay threshold, prior experience of low pay increases the probability of returning to it. Having beenlow paid at t-2 increases the probability of making the transition from higher paid at t-1 back to low pay at t: about 30%compared with about 3% if the person had also been higher paid at t-2. Looking at this the other way round, for those low paidat t-2, moving up to being higher paid at t-1 reduces the probability of being low paid at t: about 30% compared with about80% for someone who had remained low paid at t-1.
The longer the period in higher pay, the further the probability of returning to low pay falls. For someone low paid at t-4, but
71
1 The analysis described uses the first five waves of the BHPS, 1991-95. Transitions are aggregated over the waves. Earnings are adjusted to April 1997 terms.See Stewart (1998a) for details of the sample and variable condition.
higher paid from t-3 to t-1, the probability of returning to low pay at t has fallen to about 10%. This falling probability of beinglow paid, the greater the experience of higher pay is confirmed in the final block of the table. For each threshold the probabilityfalls with each additional period of higher conditioned on pay (going backwards in time).
The conclusion from Table 1 is that in aggregate terms there is considerable persistence, particularly for those who havealready been low paid for more than one period, and that the probability of becoming low paid declines the longer a person ishigher paid.
Scarring effects of low pay?
There is more than one possible explanation for this strong persistence in aggregate low pay probabilities. The dependence ofthe probability of being low paid on past low pay experience may result either from heterogeneity among individuals or fromthe impact of the experience of low pay itself. In particular, heterogeneity in characteristics which affect the probability of anindividual being low paid can create persistence in aggregate transition probabilities even if such an effect is absent inindividual transition probabilities.
Alternatively, or in addition, there may be “true”, or “structural”, dependence for individuals: being low paid in one period mayin itself increase the probability of being low paid in the next period, even relative to another individual with identicalcharacteristics who was not low paid in the first period. Employers may view low paid employment with another firm as anindicator of an individual’s low productivity and be discouraged from making a job offer. Employers may also treat holding alow paid job as a signal of a high turnover propensity. On the supply side, dependence may result from low paid employmentreducing subsequent human capital accumulation or causing the depreciation of human capital not currently being used,thereby keeping an individual’s productivity low and reducing the probability of rising out of low pay in the future. Inaddition, a spell of low paid employment may influence an individual’s perception of their market value and discourage themfrom applying for better paid jobs. Being in a low paid job may also alter worker preferences or motivation in such a way as tomake them more likely to remain in that segment of the labour market.
In all these illustrations of structural dependence, earnings correlates are altered by the experience of low pay. This contrastswith the pure heterogeneity case where individuals differ only in characteristics that affect their chances of being low paid, butthat are not affected by the experience of low pay. Persistence in the aggregate probabilities that is due to heterogeneity can beinfluenced by changing individual characteristics, e.g. by providing training, but “true” dependence may be harder to tackle.
Distinguishing between structural dependence and omitted heterogeneity is a difficult statistical problem, since the initial paystate is likely to be correlated with the omitted heterogeneity (the “initial conditions” problem). Stewart and Swaffield (1999)use a formal model for low pay transitions that allows for this endogeneity of the initial pay state. They find that (1)exogenous selection into the initial low pay state is strongly rejected, (2) ignoring the endogeneity results in considerableoverstatement of the effects of characteristics on the transition probability, and (3) there is considerable structural dependence,i.e being low paid in one period in itself increases the probability of being low paid in the next period. They estimate that thisaccounts for between 3/5 and 3/4 of the raw aggregate probability difference.
The low pay, no pay cycle
The transition probabilities looked at so far are restricted to individuals in employment at each date and therefore only givepart of the picture. As well as there being transitions between low and higher pay, there are also transitions into and out ofemployment. It is important to incorporate those not in the earnings distribution into the analysis. As well as being morelikely to remain low paid, the low paid are also more likely to move out of employment the following year. Thus restrictingattention to those who remain employees overstates the probability of the low paid moving up the earnings distribution andabove the threshold. Using the £3.50 threshold, 48% of those low paid at t-1 are still low paid at t and 17% are no longeremployees. Thus only 35% have moved up the distribution into higher pay. This probability is overstated when attention isrestricted to those who remain employed, as in Table 1.
The greater propensity of the low paid to move out of employment represents part of a cycle of low pay and no pay, which isillustrated in Table 2. First, those who are low paid at t-1 are more likely to not be working at t than those higher paid at t-1:14% against 5% in the case of the £3.50 threshold and similar differences for the other two thresholds.1 Second, those notworking at t-1 who become employees at t are more likely to be low paid than those who were already in employment at t-1:33% against 8% for the £3.50 threshold for example. Third, among those re-entrants who were not working at t-1, those whowere low paid at t-2 are more likely to be low paid again at t when they are working again than those who were higher paid att-2: 42% against 14% for the £3.50 threshold and similar differences for the other two thresholds. The low paid are more likelyto be out of work in the future; those out of work are more likely to be low paid on re-entry; and are even more likely to be soif they had been low paid prior to being out of work.
Might this cycle in aggregate probabilities be due to heterogeneity? Some light is thrown on this by the fact that this picture isrepeated within important sub-groups. There is a low pay, no pay cycle for men and women, for those aged above and below 25,and for those with and without qualifications. The picture therefore has a fractal quality: zooming in on a sub-group, one sees thesame pattern as in the original picture. The cycle does not seem to be the result of heterogeneity with respect to the main observablecharacteristics, but the possiblity that it results from heterogeneity with respect to unobservable ones cannot be ruled out.
72
1 The “not working” category includes both unemployed and out of the labour force.
Do low paid jobs act as stepping stones to higher paid jobs as is sometimes argued? This question is addressed here by lookingat those who are not employees at t-2, but are at t. Do those who enter a low paid job at t-1 have a better chance of being paidabove the threshold at t than those who remain out of employment at t-1 and enter at t? The evidence presented in Table 3indicates that they do not. For all three thresholds those who have a low paid job at t-1 have a lower probability of being paidabove the threshold at t than those not employed at t-1. This suggests that low paid jobs are more likely to act as blind alleysthan as stepping stones to positions higher up the pay distribution. This feature too is repeated within important sub-groups, bygender, age and qualifications.
Table 2: The Low Pay, No Pay Cycle Conditional Probabilities
Threshold£3.50 £4.00 £4.50
Probability of not working at t given:Low paid at t-1 .14 .12 .11Higher paid at t-1 .05 .05 .04
Probability of being low paid at t given employee at t and:Not working at t-1 .33 .47 .60Employee at t-1 .08 .15 .22
Probability of being low paid at t given employed at t, not working at t-1 and:Low paid at t-2 .42 .61 .68Higher paid at t-2 .14 .22 .32
Source: Stewart (1998a)Data: BHPS, Waves 1-5Note: Not working includes both unemployed and out of the labour force.
Table 3: Low Paid Jobs as Stepping Stones?
Probability of being higher paid at t for those employees at t, but not employees at t-2, by t-1 status
ThresholdStatus at t-1 £3.50 £4.00 £4.50
Low paid 41.9 30.5 22.9
Higher paid 87.7 84.9 87.8
Not employee 65.5 50.0 38.7
Source: Stewart (1998a)Data: BHPS, Waves 1-5
Low pay and poverty
The low pay, no pay cycle clearly has implications for poverty dynamics. The link between low pay and poverty, even in astatic context, is a complex one, depending on household composition, hours worked and the workings of the tax and benefitsystem, among other things. I start by looking at the static overlap between low pay and poverty, before looking at dynamics.1
The poverty thresholds in needs-adjusted real household net income used are given in the note to Table 4. The first two arecommonly used (closely-related definitions are used in the DSS’s HBAI statistics): (i) half of the wave 1 sample mean income,and (ii) the poorest quintile in each wave.2 The first of these is fixed (in real terms) across waves, the second varies acrosswaves. For an individual who is married, with non-working spouse and no children the three poverty thresholds correspondroughly to 40 hours at the three low pay thresholds. However for other groups the translation is very different.
The cross-sectional overlap between low pay and poverty is not very large. The first panel of Table 4 shows the percentage ofthe low paid who are in poor households.3 One in seven of those paid below the low pay threshold of £3.50 per hour are inhouseholds with equivalent net income below the first poverty threshold, rising to about a quarter for the highest of the threepoverty thresholds. Of course these figures (and those for the other thresholds) are all much higher than the corresponding
73
1 This section is based on Stewart (1998b).2 Calculated across full samples, including those not employed and the retired.3 Since the focus here is on the interplay with low pay, pensioners and those under 18 are excluded. 11.0%, 13.5% and 17.4% of remaining individuals are in
households with adjusted net income below the three thresholds respectively.
figures for those paid above the threshold - between five and ten times as great in all cases.4 However it is still the case that thevast majority of those who are low paid are not in poor households.
Typically poor households do not contain earners. The third panel of Table 4 indicates that between three-quarters and four-fifths ofpre-retirement adults in poor households are not employees. Looking at all those in poor households relatively few are low paid,because relatively few are in employment. If we restrict attention to those in work, the association is much closer. A substantialnumber of the employees in poor households are low paid, particularly for the higher two low pay thresholds. For example, overthree-quarters of employees in households with income below the first poverty threshold earn below £4.50 per hour.
Table 4: The Overlap between Low Pay and Poverty
Poverty Threshold(1) (2) (3)
Percent of low paid who are in poor households:LP threshold = £3.50 14.5 18.2 25.5
£4.00 12.3 15.7 22.5£4.50 10.5 13.7 19.8
Percent of higher paid who are in poor households:LP threshold = £3.50 1.9 2.8 4.6
£4.00 1.4 2.1 3.7£4.50 0.9 1.5 2.8
Percent of pre-retirement adults in poor households who are employees:19.0 21.0 25.5
Percent of employees in poor households who are low paid:LP threshold = £3.50 42.7 38.5 34.6
£4.00 61.2 56.0 51.5£4.50 76.4 71.6 66.4
Percent of “permanent” low paid who are in poor households:LP threshold = £3.50 49.4 52.5 63.1
£4.00 42.1 46.5 58.6£4.50 41.7 45.6 56.1
Percent of “temporary” low paid who are in poor households:LP threshold = £3.50 29.4 34.7 43.4
£4.00 25.3 29.9 36.6£4.50 18.1 21.5 29.3
Percent of never low paid who are in poor households:LP threshold = £3.50 10.4 12.2 16.0
£4.00 8.8 10.3 13.6£4.50 7.2 8.7 11.2
Source: Stewart (1998b) Data: BHPS, Waves 1-4Notes:Poverty thresholds (in April 1997 terms):(1) £133.08/week in all 4 waves: half wave 1 mean.(2) £139.70/week in wave 1 to £147.14/week in wave 4: poorest quintile in each wave.(3) £160/week in all 4 waves.
Whilst, as stated above, the overlap between low pay and poverty for the full sample is not very great, it has increasedconsiderably in the last twenty years or so. The poor in the 1990s are more likely to be working than the poor in the 1970s.This rise in the number of working poor over this period is all the more remarkable because it has occurred despite another ofthe most important trends of the period: the fall in the proportion of families which have at least one worker.
An important consideration when looking at the overlap between low pay and poverty is transitory variation. There aresignificant differences between those who are “permanent” low paid and those who are “temporary” low paid. The importanceof this for the overlap between low pay and poverty is reinforced by the cycle of low pay and no pay.
Due to this cycle of low pay and no pay some care has to be taken in making the distinction operational. Of those low paid in atleast one of the 5 waves of the BHPS used in the last section, around half were also not employed in at least one wave. The“permanent” low pay group is defined here to be those who were low paid in at least one wave and higher paid in none, i.e. in the
74
4 In contrast they are much lower than the corresponding figures for those who are not working: 31.8%, 38.0% and 45.9%.
waves when they were not low paid they were out of a job. The “temporary” low paid are those who were low paid in at least onewave and also higher paid in at least one wave. Finally the never low paid group are those who are higher paid in at least onewave and low paid in none. About 1/3 of those low paid in at least one wave are permanent low paid under this definition. Thepermanent low pay proportion is about 2/3 of the corresponding cross-sectional low pay proportion for each threshold.
Table 5: Movements into and out of Low Pay and Poverty
Poverty Threshold(1) (2) (3)
LP threshold = £3.50
Probability move into poverty, given:move from higher to low pay 11.4 13.6 13.5remain higher paid 1.0 1.2 1.7
Probability move out of poverty, given:move out of low pay 85.4 74.4 68.4remain low paid 60.1 63.4 49.9
LP threshold = £4.00
Probability move into poverty, given:move from higher to low pay 11.2 11.2 11.5remain higher paid 0.6 0.8 1.3
Probability move out of poverty, given:move out of low pay 88.8 74.2 64.7remain low paid 61.1 57.8 48.8
LP threshold = £4.50
Probability move into poverty, given:move from higher to low pay 8.2 8.5 9.3remain higher paid 0.3 0.5 1.0
Probability move out of poverty, given:move out of low pay 87.1 73.2 66.1remain low paid 63.9 58.6 49.8
Source: Stewart (1998b)Data: BHPS, Waves 1-4Note: Poverty thresholds as in notes to Table 4.
Around half of those who were “permanent” low paid using the £3.50 threshold are in a poor household for at least part of theperiod if the lowest poverty threshold is used. (Poverty in at least one of the years seems the appropriate focus from a welfareperspective due to the borrowing constraints faced by poor households.) In contrast to this, only 1 in 10 of those who werenever low paid (using the £3.50 threshold) are in poverty in any of the waves if the lowest threshold is used, around 1 in 6 ifthe highest threshold is used.1 The position for those who were “temporary” low paid is roughly half way between that for the“permanent” and never low paid. Clearly from this more longitudinal (although still relatively short-run) perspective, theoverlap between low pay and poverty is considerably greater than when viewed in a single-point snapshot context.
It is also informative to look at the degree of association between movements into and out of poverty and movements into andout of low pay. Table 5 first considers the association between movements into low pay and movements into poverty. In thiscase the group in focus is those who were paid above the low pay threshold and were not in poor households at t-1. For thosewho became low paid at t, 11.4% move into poverty, compared with only 1.0% among those who do not become low paid (firstpoverty threshold and £3.50 low pay threshold). Thus becoming low paid has a large (relative) impact on the probability of thehousehold becoming poor. The other thresholds show similar differences between the two groups.
Next consider the association between movements out of low pay and movements out of poverty. Now the group in focus isthose who were both low paid and in a poor household at t-1. For those who move out of low pay at t, 85.4% move out ofpoverty, compared with 60.1% among those who remain low paid. Whilst this indicates that leaving low pay is not the maindeterminant of leaving poverty, it is also clear that it is a very important influence.
To sum up this section, the snapshot overlap between low pay and poverty is not large, but when one takes a more inter-temporal view, the association is much more important.
75
1 Those out of work in all four waves being analysed are excluded.
Conclusions
The main findings reported in this paper are as follows. There is considerable persistence in low pay, in aggregate terms,particularly for those who have already been low paid for more than one period. In addition, the probability of becoming lowpaid declines the longer a person is higher paid. Much of the persistence in low pay is “true” or “structural” dependence ratherthan merely the result of heterogeneity.
There is strong evidence of a cycle of low pay and no pay. The low paid are more likely to be out of work in the future; thoseout of work are more likely to be low paid on re-entry; and are even more likely to be so if they had been low paid prior tobeing out of work.
The hypothesis that low paid jobs act as stepping stones to higher paid jobs is not supported by the data. The evidencepresented here suggests that low paid jobs are more likely to act as blind alleys than as stepping stones to positions higher upthe pay distribution.
The cross-sectional overlap between low pay and household poverty is not large. However around half of those classified as“permanent” low paid are in household poverty. From this more longitudinal (although still relatively short-run) perspective,the overlap between low pay and poverty is considerably greater than when viewed in a snapshot context. It is also found thatmovements into and out of low pay exert an important influence on movements into and out of poverty.
ReferencesStewart, Mark B. (1998a), “Low Pay in Britain: Piecing Together the Picture”, mimeo, University of Warwick.Stewart, Mark B. (1998b), “The Dynamics of the Relationship Between Low Pay and Poverty in Britain”, mimeo, University of Warwick.Stewart, Mark B. and Joanna K. Swaffield (1999), “Low Pay Dynamics and Transition Probabilities”, Economica, vol. 66, pp.23-42
76
‘Wage mobility in Great Britain’
Richard Dickens, Research Fellow, Centre for Economic Performance, LSE
Introduction
The sharp increase in cross section wage inequality that occurred in the UK since the late 1970’s has been well documentedand the potential causes have been extensively researched (See Machin, 1996). However, relatively little attention has beengiven to the important question of how much workers move within the wage distribution from one period to the next. Crosssection studies provide only a snapshot of the earnings distribution, so they only tell us about differences between workers at apoint in time. We know that the distribution of this snapshot has widened considerably but this tells us nothing about earningsdifferences between workers over their “lifetimes”. If workers experience a high degree of mobility in the wage distributionfrom one period to the next then this will act to offset some of the cross sectional differences; the low paid today could well bethe high paid next year and vice versa. In addition, it is possible that mobility has increased over the last couple of decades tooffset some or all of the rise in cross sectional dispersion, so that “lifetime” earnings differences are unchanged. Alternatively,it could be the case that mobility has fallen which would exacerbate the increase in cross sectional dispersion.
Until recently little was known about these issues. However, with the advent of a number of longitudinal data sources, thatfollow the same individuals over time, evidence on earnings dynamics and mobility is becoming more widespread.1 In thischapter I provide an overview of this new evidence.
What do we know about Earnings Mobility?
• Mobility in the wage distribution over the course of one year is limited and those individuals that do move tend to experiencequite short range mobility.
Tables 1a and 1b (from Dickens, 1997) present information on transitions of males and females between 1993 and 1994 fromthe New Earnings Survey. The table includes information about transitions within the (hourly) wage distribution and transitionsinto and out of other labour market states. The figures reported are the percent of individuals in a certain decile or labourmarket state in 1993 who are in a given decile or labour market state in 1994.2
77
1 See Atkinson, Bourgignon and Morrison (1992) for a survey of the earnings dynamics literature at that date. Stewart and Swaffield (1997), Gosling, Johnson,McCrae and Paull (1997) and Ball and Marland (1996) all provide an analysis of earnings mobility in Great Britain.
2 Earnings are split into deciles; ten bands each containing 10% of employees. The 1st decile contains the lowest paid 10%, the 2nd decile contains the nextlowest paid 10% and so on up to the 10th decile, which contains the highest paid 10% of workers.
Notice the high degree of persistence associated with labour market states and deciles of the wage distribution. The diagonalelements are all much higher than the off diagonal elements. For example, over 48% of males in the bottom decile in 1993 arestill there one year later. Many are leaving employment altogether for unemployment or inactivity and some have missingwage details the next year. Consequently, only 20% move up the wage distribution. Most of those that do move up, do not risevery far. Two thirds only make it to the 2nd decile and very few progress beyond the median.
There is more movement within the middle deciles of the distribution with about 40% remaining in their decile of origin.
Table 1a: Male One Year Transition Rates (NES) 1993/94. Percent of Given State in 1993 in Given State in 1994.
Notes: 1. The NES is conducted in April.2. The unemployed are those claiming unemployment related benefits.3. The missing includes the self employed, retired, inactive and those not captured by the survey.
Again, mobility tends to be quite short range with a high concentration around the deciles close to the diagonal; 68% ofthose starting in the 5th decile either stay there or move to an adjacent decile. Mobility is much lower at the top of thedistribution; 70% of those in the 10th decile retaining their status. Those that do move down from there tend not to movevery far.
Table 1b presents the same transition matrix for women. A similar pattern emerges, with evidence of immobility in earningsand labour market state. There is some evidence that women face slightly higher levels of mobility than men, with lessemployees remaining in the same decile year on year.
• Workers in the bottom end of the wage distribution are much more likely to exit into unemployment or inactivity.Similarly those unemployed and inactive that do find work tend to enter into low paying jobs.
We see that far more of those who leave employment for unemployment (or drop out) do so from the bottom end of thewage distribution than elsewhere. For example, 6.4% of the 1st decile are unemployed a year later compared to 2.8% of the5th decile and 1.8% of the top decile. Males in the bottom decile are twice as likely to become unemployed than those inthe 5th decile. The unemployed face a high degree of persistence, with 55% remaining unemployed a year later. Those thatdo make it into work are much more likely to do so at low wages. Some 4.6% enter at the bottom decile compared to 0.9%in the 5th decile and practically nobody at the top.3
• Earnings mobility is greater when measured over a longer time period (5 years or 14 years) but even here there is littlelong range movement.
Table 2 presents five year decile transition matrices of hourly earnings for males between 1989 and 1994 from the NES(See Dickens, 1997). As expected mobility is higher when measured over this longer time span with much less
78
Notes: See Table 1a.
Notes: See Table 1a.
Table 1b: Female One Year Transition Rates (NES) 1993/94. Percent of Given State in 1993 in Given State in 1994.
Table 2: Male Five Year Transition Rates (NES) 1989/94. Percent of Given State in 1989 in Given State in 1994.
3 This is in line with the existence of a low pay-no pay cycle as demonstrated by Mark Stewart in this volume.
concentration on the diagonals than over one year; only 22% of males remain in the bottom decile after five years.Nevertheless many of the employed men in the bottom decile in 1989 have dropped out of employment altogether by 1994 sothat only 30% have actually moved up the wage distribution, compared to 20% moving up over one year. Amongst those thatdo move few progress very far, with two thirds finding themselves in the next two deciles. There is more mobility from themiddle deciles, but most movement is still fairly short range. Once again those in the top decile seem to retain their status themost, with some 48% of males remaining there after five years and with very few experiencing large wage falls. The figuresfor females (See Dickens, 1997) display a very similar pattern to these, with perhaps some evidence of slightly more mobility.
Ball and Marland (1996) examine earnings mobility of males over a fourteen year time span (1978/79 to 1992/93) using theLifetime Labour Markets Dataset.4 Over this long time span they find more mobility but much of this is still quite short range,with most workers only moving a couple of deciles. Many of those in the bottom part of the wage distribution fall out ofemployment onto benefits. They also find evidence of strong persistence for a core group of workers. For example, 13% ofyoung workers who were in the bottom decile in 1978/79 are still there in 1992/93. However, some 25% of these are in thebottom decile for every intervening year. Their results suggest that for some individuals low pay and benefit dependency is avery persistent phenomenon.
• Short run mobility rates have fallen for men and women since the late 1970s which has exacerbated the rise in cross sectionwage inequality.
Analysing changes in the transition matrices between the late 1970s and late 1980s one finds an increase in the proportion ofemployees who remain in the same decile from one year to the next (See Dickens, 1997). The results suggest a slightly largerfall in mobility for males than females. In addition, examining five year transition matrices one also find a fall in mobility formales but little change for females.
However, there are some potential problems with the analysis of changes in mobility using transition matrices. Whether anindividual crosses a decile threshold depends on the wage range within each decile. We know that this range has risen due tothe rise in cross section dispersion so that movements across deciles are not comparable over time.
In order to resolve this problem I employ an alternative measure of mobility based on the individual’s actual position withinthe wage distribution and examine changes in this between two time periods. I do this by computing each individual’spercentile ranking based on their wage in a given year. This gives a measure from zero to one for each individual depending onwhere they lie in the wage distribution.
Figure 1a plots the percentile ranking of individuals in 1978 against that in 1977. If there is no mobility then one would see apositively sloped diagonal line, since nobody changes position in the wage distribution from one year to the next. Ifindividual’s earnings are uncorrelated between the two years then one would see a random scattering. If individual’s earningsare negatively correlated between the two years, so the lowest paid became the highest paid and vice versa, then one wouldobserve a negatively slopped diagonal line starting from the top left corner of the figure. In fact we see a concentration ofindividuals around the positively slopped diagonal. This concentration is much higher at the top and bottom of the distribution,indicating lower levels of mobility at these points. Figure 1b presents the same scatter plot for earnings rankings in 1988 and1989. It is clear that there is a greater concentration around the diagonal in this latter year, providing evidence that mobility hasfallen over this time period for males.
These ranking plots for females (not presented) provide a similar pattern of changes in mobility over this time period.
I have also computed an index of mobility based on these changes in percentile rankings of individual’s earnings (See Dickens,1997, for more details). The index is plotted for males and females from 1976 to 1994 in Figure 2. This index takes a minimumvalue of zero when there is no mobility, a maximum value of one when earnings are perfectly negatively correlated and a value
79
Figure 1a Earnings Ranking in 1977 and 1978:Males
Figure 1b Earnings Ranking in 1988 and1989: Males
4 See also the chapter by Rebecca Endean in this volume.
of two thirds when earnings are uncorrelated across the two years. Taking the period as a whole the indices fall. By 1994 theyhave fallen by 41% for males and females since peaking in 1980.5
• There is some weak evidence that long run mobility rates have fallen for men but not for women since the late 1970’s.
One can also use this ranking methodology to examine changes in longer term mobility. In Dickens (1997) I compute thismobility index for males and females over a five year time span. After controlling for the effects of inflation I find some weakevidence that mobility has fallen for males but no evidence of any change for females.
• At least half of the rise in cross section inequality can be explained by increasing permanent differences between workers.
Another way of analysing this issue is to decompose earnings differences into that part which is permanent and that part whichis transitory and examine how these have changed over this time period. Permanent differences between individuals are thosethat last forever and may reflect the effects of fixed unobserved ability differences. Transitory differences die out over time andmay reflect the effects of job characteristics, bonus payments, etc. In Dickens (1998), I find that about half of the rise isaccounted for by an increase in permanent differences, with the other half accounted for by a rise in (highly persistent)transitory differences. Consequently, over half of the rise in cross sectional earnings inequality is explained by increasingdifferences between individuals that last for a good many years.
Conclusions
In this chapter I have reviewed the evidence that has emerged over the last couple of years on the dynamics of earnings. Thishas shown that the picture that has sometimes been painted of a mobile society is far from the truth. In fact, the evidence showsa high degree of immobility with little long range wage movements. In addition to this there is evidence showing that earningsmobility has fallen since the late 1970’s. Given that we have also seen a sharp rise in cross sectional wage inequality over thistime period, this tells us that not only has the gap between the rich and poor risen but the ability of the low paid to close thisgap has fallen considerably. Far from offsetting the increase in cross section wage inequality, changes in mobility appears tohave exacerbated this rise.
The evidence on wage mobility suggests that policies that raise the earnings of the poorly paid will provide some welcomerelief. For example, the minimum wage will help to compress the pay distribution. In addition, mobility both within and intoand out of the wage distribution means that more people will be affected by the minimum wage than suggested by cross sectionfigures. However, it is unlikely to do much to help individuals progression up the wage distribution. Similarly, policies that aimto get people back into work must not then leave the individual to their own devices once they are in work. Individuals placedin poorly paid entry jobs with little prospects are unlikely to progress on from these.
ReferencesAtkinson, A, F. Bourguignon and C. Morrisson, (1992) Empirical Studies of Earnings Mobility. Harwood Academic Publishers, Reading, 1992.Ball, J and M. Marland (1996) “Male Earnings Mobility in the Lifetime Labour Market Database”, Department of Social Security, Analytical ServicesDivision Working Paper No 1.Dickens, R. (1997) “Caught in a Trap? Wage Mobility in Great Britain: 1975-94” Centre for Economic Performance Discussion Paper No. 365.Dickens, R. (1998) “The Evolution of Individual Male Wages in Great Britain: 1975-95” Centre for Economic Performance Mimeo.Gosling, A., P. Johnson, J. McCrae and G. Paull. (1997) The Dynamics of Low Pay and Unemployment in Early 1990s Britain, Institute for Fiscal Studies,London.Machin, S. (1996) “Wage Inequality in the UIC”, Oxford Review of Economic Policy, Vol 12, No. 1, pp47-64.OECD (1997) “Earnings Mobility: Taking a Longer Run View”, Employment Outlook, July 1997, pp27-61.Stewart, M.B. and Swaffield, J. (1997) “The Dynamics of Low Pay in Britain”, in Gregg, P. (ed) Jobs, Wages and Poverty: Patterns of Persistence and Mobilityin the New Flexible Labour Market, Centre for Economic Performance, pp36-51.
80
Figure 2 One Year Mobility Index: Males and Females 1976-94
5 There is a potential problem arising from the fact that mobility measures in the NES are highly correlated with inflation (See Dickens, 1997). Nevertheless, Istill find evidence of a 22% fall in this mobility index for males and an 11% fall for females between 1976 and 1994 after controlling for the effects ofinflation.
‘Patterns of work, low pay and poverty – evidence from the Lifetime Labour Market Database andthe British Household Panel Survey’
Rebecca Endean, Analytical Services Division, Department of Social Security
Introduction
This chapter falls into two parts. The first section presents some descriptive statistics based on the Lifetime Labour MarketDatabase (LLMDB) showing patterns of low pay, high pay and no pay in the early 1980s and early 1990s (see Nicholls,Marland and Ball, 1997). The aim is look at the extent of persistence in various labour market states for working ageindividuals and whether this has changed over time. As such this work complements the chapters by Richard Dickens and byMark Stewart which look at similar patterns but use different data sources. The second section looks at patterns of low pay,high pay and no pay for working age individuals in the British Household Panel Survey (BHPS) in the early 1990s and relatesthis to patterns of household income. This brings together the work on household income dynamics – see Stephen Jenkins’chapter – which work on individual labour market dynamics.
The Lifetime Labour Market Database (LLMDB)
The LLMDB is a 1 per cent sample of all National Insurance records. Each individual, aged over 16, has a unique NationalInsurance number and, as the same individuals are sampled each year, the information can be linked over time to provide alongitudinal labour market information. The following information, held on the database, has been used in this analysis:
• personal information: the database contains information on age, sex, death and migrant status;
• National Insurance contributions: all classes and amounts of contributions are held. Contributions are paid by employeesand the self employed. People receiving certain types of Social Security benefits are credited with contributions.Individuals who have the contribution records protected because of their caring responsibilities (Home ResponsibilitiesProtection) can also be identified;
• earnings information: total annual taxable earnings for each person for whom employers provide an end-of-year PAYEreturn.
There are some gaps in the data:
• the information on contributions and credits can be expected to be more or less complete. The coverage of the informationon earnings is less certain. Employers are legally required to provide information on employees who earn above theNational Insurance Lower Earnings Limit but the coverage of people who earn below the LEL is unlikely to be complete;
• there is no earnings information for periods of self-employment but it is possible to identify periods of self-employmentwhere Class 2 National Insurance Contributions are paid;
• there is no information about household characteristics or other sources of income. As a result it is not possible to look attrends in the extent of low-income persistence.
For employees, the database records annual gross earnings. As a result people who work fewer hours per week or fewer weeksper year than other people earning the same wage will be located further down the earnings distribution. As a result thefollowing analysis is limited in what it can tell us about changes in labour market dynamics. However, it is useful in that itprovides a dynamic picture of patterns in individuals’ income from employment.
There are gaps in people’s contribution records – years where we have no information on either earnings or creditedcontributions or periods of Home Responsibilities Protection. These gaps can occur because the individual was out of work butnot claiming the relevant benefits or in work and not paying National Insurance. In the following these gaps are assumed to beperiods of no pay. This is a strong assumption, as some of the gaps will refer to periods when the individual was out of thecountry1 or in an institution (prison, hospital etc). The main alternative assumption, which is to reject all cases with gaps intheir contribution records, would lead to the sample becoming biased and the results unrepresentative of the working agepopulation as a whole. However, the estimates presented here were also calculated using this alternative way of defining thesample, and the main trends over time shown below were also evident in this restricted sample, although the numbers fallinginto each category were slightly different.
The LLMDB currently holds information for the 1978/79 to the 1995/96 period. The following analysis takes two five-yearperiods or “windows”, 1979/80 to 1983/84 inclusive and 1990/91 to 1994/95 inclusive. These periods relate to broadly similar
81
1 However some attempt was made to remove some of the National Insurance numbers which refer to people who are no longer in the country. Excluded fromthe sample were any National Insurance records where there were no recorded contributions in any year in the relevant five-year period. This will rejectindividuals who left the country before the beginning of the relevant period. Also excluded were people for whom there is a record on the file denoting thatthey migrated into the country after the start of relevant period. After these restrictions were made the resulting number of cases matched well with theestimated resident UK population in the relevant periods.
points of the economic cycle and so should not be affected much by cyclical changes. The sample was restricted to thosebelow State Pension Age at the beginning and end of the period i.e. men aged 16-59 in 1979 and 1990 and women aged 16-54in 1979 and 1990. In these five year periods the proportions of people falling into each of the following eight categories wereestimated:
• High pay in every year;
• Low pay in every year;
• No pay in every year;
• Either no pay or low pay in every year;
• Either high pay or low pay in every year;
• Either no pay or high pay in every year;
• At least one period of self employment during the five year period; and
• All other combinations.
People were classed as self employed in a year if they had some Class 2 National Insurance Contributions at any point duringthe year. People were classed as having no pay if they had no pay as employee recorded and no Class 2 National Insurancecontributions. This category will include some people who were working during the year but had very low earnings – belowthe LEL for employees and the small earnings limit for the self employed.
In order to decide whether people were high paid or low paid two arbitrary earnings thresholds where chosen:
• A threshold which was held constant in real terms over the period. This was set at just under £3000 per annum in 1979/80rising to £7000 per annum in 1994/95 (current prices). People with annual earnings below this threshold were categorisedas low paid, above the threshold as high paid.
• A threshold which increased at the same rates as average earnings over the period. This was set at just under £3000 perannum in 1979/80 rising to £9500 in 1994/95.
These thresholds are completely arbitrary and one useful extension of this work would be to see whether the results aresensitive to changes in the threshold.
Absolute Pay Threshold
Table 1 shows the proportions of the sample falling into each of the eight categories specified above where low pay is definedusing the absolute threshold. The main points to note are:
• Half the sample fell into the three “permanent states” – high pay in every year, low pay in every year and no pay in everyyear – in both periods. This demonstrates considerable persistence within the labour market. If the population wererandomly distributed between the states in each year (with the same propensity as observed in the population as a whole)then the numbers falling into these first three categories would be very small. Another way of looking at this is to calculatewhat proportion of those in each state remained there for all of the following four years. 69 per cent of those with high payin 1990/91 had high pay in all of the next four years, 54 per cent of those with no pay in 1990/91 had no pay in all of thenext four years and 29 per cent of those with low pay in 1990/91 had low pay in all of next four years;
• Men and women had significantly different labour market patterns. Men are more likely to have had high pay in every yearand less likely to have had low pay in every year, no pay in every year and either low pay or no pay than women are;
• 14 per cent of the population had either low pay or no pay in every year compared with 5 per cent of the population havingeither high pay or no pay in every year and 13 per cent having either high pay or low pay in every year. This indicates thatpeople without work are more likely to combine spells out of work with spells of low pay than with spells of high pay. Andthat people in low pay are just as likely to combine spells of low pay with spells of no pay than with spells of high pay eventhough a much higher proportion of the working age population is high paid than out of work. These patterns providefurther supporting evidence of a “low pay no pay” cycle as demonstrated by Stewart and Swaffield;
• Using this absolute pay threshold then, for the population as a whole, the proportions falling into each category were similarin 1979/1983 and 1983/1994;
• For men there has been a decline in the proportion with high pay in every year, an increase in the proportion with no pay inevery year and a small rise in the proportion with either no pay or low pay in every year. The reverse is true for women,there has been an increase in the proportion with high pay in every year, and a fall in the proportion with low pay every
82
year, no pay in every year and either low pay or no pay in every year.
Tables 2 and 3 show the position for men and women by age group. The main points to note are:
• The increase in the proportion of men with no pay in every years is largest (in percentage point terms) for older men – theproportion with no pay in each of the five years increased from 9 per cent to 18 per cent. The fall in the proportion of menwith high pay in every year was also sharpest for this age group. However there was still a significant fall in the proportionof prime age men with high pay in every year from 54 per cent to 48 per cent;
• For women, the rise in the proportion of women with high pay in every year was largest for prime aged women up from 13per cent to 24 per cent.
Relative pay threshold
Tables 4,5,6 present the same estimates using the threshold that increases with average earnings over time. The main points tonote are:
• Unlike tables 1,2,3, which use the absolute pay threshold, there is a decrease in the proportion of the population who hadhigh pay in every year and a corresponding increase in the proportion of the population with low pay and either low pay orno pay in every year from 15 per cent to 17 per cent. This reflects the widening of the distribution of earnings over thisperiod and provides some evidence that the low pay-no pay cycle has increased (providing low pay is defined in relativeterms);
• There is no increase in the proportion of the population experiencing either high pay or low pay in every year. Thereforethese estimates do not suggest that the increase in the numbers of people who are relatively low paid at a point in time wereoffset by increased mobility within the earnings distribution;
• For men there is a significant increase in the proportion with either low pay or no pay in every year up from 6 per cent to 10per cent. For women there is only a slight decline in the proportion falling into this category;
• For women the increase in the proportion with high pay in every year is now slightly smaller than the increase in theproportion with low pay in every year.
British Household Panel Survey
This section presents information on the links between individual labour market patterns and household incomes. Differencescan occur between an individual’s position in the labour market and her position within the household distribution of incomebecause of:
• Household structure, the number of adults and children in the household;
• Earnings of other members of household;
• Other income sources, benefits, investment income;
• The structure of taxation.
As a result although income from employment is an important determinant of household income these other factors also play apart in determining who lives in a low-income household. In order to explore this the BHPS was used to look at working agemen and women, their employment patterns and household income patterns over the 1991-1995 period. Individuals’employment patterns were classed as:
• Favourable – high pay in every year using the same absolute pay threshold as in the LLMDB analysis but with monthlygross pay rather than annual pay as the relevant pay variable;
• Unfavourable – either no pay in every year or low pay in every year or a combination of low pay or no pay in every year;
• Self employed – at least one year our of the five spent in self-employment;
• Mixed – all other employment patterns.
Then people were separated into those who lived in households with persistently low income over the 1991 to 1995 period andthose who did not. Persistently low income was defined as living in a household which was in the bottom 30 per cent of theincome distribution in at least three years out of five and no more than one year in the top 60 per cent of the incomedistribution. (For more detail on the rational behind this classification and the income variable used see DSS (1998) HBAI1996/7 chapter 4.)
83
84
Table 7 shows the proportion of all working age individuals who lived in households with persistently low incomes byemployment pattern. 13 per cent of working age individuals fell into the persistently low income category with slightly moreworking age women than men. The proportion of the overall population with persistently low incomes is higher becausepensioners and children are more likely to live in such households than working age people. Working age individuals withunfavourable employment patterns are more likely to live in households with persistently low incomes than other working ageindividuals are. 38 per cent of working age men and 27 per cent of working age women with these unfavourable employmenthistories fell into this position. However, the majority of these people did not live in households with persistently low incomessuggesting that in some of these cases other factors – such as the existence of other earners within the households – may behaving an impact on household incomes.
Table 8 shows that the majority of individuals living in households with persistently low incomes had unfavourableemployment histories. 67 per cent of those living in households with persistently low household incomes also hadunfavourable employment histories. The proportion of low-income women with unfavourable employment histories – 86 percent – was higher than the proportion of low-income men – 44 per cent.
Conclusion
The results presented in this chapter provide some evidence of the extent of labour market persistence and the links betweenindividual employment patterns and household incomes. However, descriptive statistics can only be indicative of causation andmore research work is needed using longitudinal data to explore the factors driving these changes and the implications forpolicy. However, in summary, the main points to be noted are:
• Men’s labour market patterns worsened between the two periods. They became more likely to be continuously out of work,continuously low paid or in a low pay-no pay cycle – where low pay is defined using a relative threshold. The change wasmost dramatic for older men but there were also signs of a decline in the position of prime age men;
• Women’s employment patterns improved over the period relative to men’s but they are still more likely to be continuouslyout of work, continuously low paid or in a low pay-no pay cycle than men are;
• This work does not suggest that the increase in the number of people with relatively low pay over the period was offset byan increase in the amount of mobility between low pay and high pay;
• Most working age women with persistently low household incomes also had unfavourable employment patterns; for men the pattern was less clear cut.
Table 1: Patterns of low pay, high pay and not pay 1979/1983 and 1990/1994using a low pay threshold fixed in real terms; all of working age.
Column Percentages
Table 2: Patterns of low pay, high pay and no pay 1979/1983 and 1990/1994using a low pay threshold fixed in real terms; men by age group
Column Percentages
Table 3: Patterns of low pay, high pay and no pay 1979/1983 and 1990/1994using a low pay threshold fixed in real terms; women by age group
Column Percentages
Table 4: Patterns of low pay, high pay and no86pay 1979/1983 and 1990/1994 using a low pay threshold fixed in real terms;
All of working age.
Column Percentages
85
Total
Table 7: Working age individuals with different employment patterns over the 1991-95 periodby position in the household income distribution
Row percentages
Notes: The variables are defined as follows:
Favourable employment patterns – High pay in every year. Unfavourable employment patterns – Low pay in every year, or no pay in every year, or every year spent in either low pay orno pay.Mixed employment patterns – All other employment patternsSelf employed – At least one year when self employed
Persistent low income – At least three years out of 5 in the bottom 30 per cent of the income distribution and no more than oneyear in the top 60 per cent of the income distributionOther income patterns – All other patterns
Source: HBAI (BHPS)
Table 5: Patterns of low pay, high pay and no pay 1979/1983 and 1990/1994using a low pay threshold fixed in real terms; Men by age group
Column Percentages
Table 6: Patterns of low pay, high pay and no pay 1979/1983 and 1990/1994using a low pay threshold fixed in real terms; Women by age group
86
Table 8: Working age individuals with different employment patterns over the 1991-95period by position in the household income distribution
Column Percentages
Notes: As table 7
Source: HBAI (BHPS)
Column PercentagesReferences Department of Social Security (1998). Households Below Average Income 1979-1996/97, Corporate Document Services, London.Dickens R. (1997). “Male Wage Inequality in Great Britain: Permanent Divergence or Temporary Differences” in “Jobs, Wages and Poverty: Patterns ofPersistence and Mobility in the New Flexible Labour Market”, ed Paul Gregg. Centre for Economic Performance.Jarvis S. and Jenkins S.P. (1997). ‘Low income dynamics in 1990s Britain’, Fiscal Studies 18: 1-20.Jarvis S. and Jenkins S.P. (1998). ‘How much income mobility is there in Britain?’, Economic Journal 108: 428-443.Jenkins S.P. (1998). ‘Modelling household income dynamics’, Presidential Address to the European Society for Population Economics, Amsterdam, June1998, unpublished paper, University of Essex.Nicholls M., Marland M. and Ball J. (1997) “The Department of Social Security’s Lifetime Labour Market Database” in “Jobs, Wages and Poverty: Patterns ofPersistence and Mobility in the New Flexible Labour Market”, ed Paul Gregg. Centre for Economic Performance.Stewart M. and Swaffield J. (1997). ‘The Dynamics of Low Pay in Britain” in “Jobs, Wages and Poverty: Patterns of Persistence and Mobility in the NewFlexible Labour Market”, ed Paul Gregg. Centre for Economic Performance.
87
88
‘The Impact of Unemployment and Job Loss on Future Earnings’
Paul Gregg, London School of Economics
Introduction
The growing evidence on poverty dynamics highlights the important role for variation in access to work and wages. Wagedynamics and low pay in general are considered by Mark Stewart and Richard Dickens elsewhere in this volume and there is astrong link between this paper and their work on the low pay – no pay cycle. This paper, however, looks at the availableevidence on:
• the extent of and consequences of involuntary job loss, plus the nature of jobs and wages for returners to the labour market;
• the evidence on durations of joblessness and the scarring effects of long-term unemployment (duration dependence);
• the evidence on concentration of unemployment over long periods and thus including repeat spells. Plus the evidence thatunemployment now increases the future risk of unemployment – that there is causal relationship behind repeat spells.
This represents rather a lot to get through so it is brief and in places patchy. Documenting the evidence of the issues ofworklessness is relatively straight forward but interpretation is far more difficult. This requires disentangling what is due to theindividual and that due to area variation in the availability of work. And where it is down to the individual, whether it is due toheterogeneity between people that develops before entry into the labour market (including innate variation), and that which hasorigins in the shocks and events that occur through peoples working lives. This is important to assess whether, when and howto intervene.
Job loss
The key points are:
• 1.8 million workers in Britain per annum will suffer involuntary job loss. This varies from about 1.4 million to 2.2 millioncounter-cyclically;
• on average a worker re-entering work after involuntary job loss will earn 9% less than in the previous job;
• when compared with an employee in continuous work the average wage loss of someone reentering the workforce is 14%;
• the cost of job loss is most severe among more experienced workers and those with longer tenure. Length of time beforeregaining work is also a key predictor of the wage gap;
• tenure is the most important single predictor of job loss. There is an 11.7% chance of a worker with tenure less than 1 yearlosing a job, this drops to just 4% after 5 years;
• about 7% of the wages declines associated with job loss are still present after 2 years of employment. US evidence suggestslonger run persistence is due to repeat job loss. In the UK there are certainly far more temporary and part-time jobs in thestock of entry jobs than amongst all jobs;
• wages among entry jobs – those taken by those out of work have been declining relative to other jobs since 1980. In the lastfew years they have been declining in real terms.
89
Table 1
Annual separation and displacement rates
Year Separation rate Displacement rate
90/91 24.7 7.891/92 19.8 6.992/93 20.5 7.293/94 21.9 6.494/95 20.6 5.795/96 26.8 5.1
Average 22.4 6.5
Source: BHPS
90
Incidence of displacement
Industrial restructuring, changes in technology, industrial location and recession are all associated with worker displacement: theinvoluntary separation from a job. Who is at risk from displacement? Around one in five employees, some 5 million workers,will separate from their jobs during one year (Table 1). On average, some 6.5% of employees will lose their jobs as the result ofdisplacement. This varies across the cycle from around 1.4 million in a good year to 2.2 million in a recession year.
The most important single predictor of displacement is job tenure. There is an 11.7% chance that a worker in a job for less than12 months will lose a job, compared with a less than 4% chance of displacement for a worker in a job for 5 years or more,(Table 2). Men are almost twice as likely to be displaced as women, 8.4% compared to 4.5%. The displacement rate foryounger workers, under the age of 25, is 11.5%, twice that of other age groups.
There is some evidence that education affects displacement. The lower the level of educational attainment, the higher thedisplacement rate. However, the difference between the highest and lowest education groups, at around 2 percentage points, isnot large when compared with other variation. The industrial sector with the highest displacement rate is construction, at13.4%, The lowest is the public services at around 2%. There is little variation across other sectors. Most of those displaced get
Table 2
Separation and displacement rates by worker and firm characteristics
Separation rate Displacement rate
Gender
Female 19.2 4.5
Male 23.4 8.4
Age
Youths <25 36.6 11.4
Prime 25-49 19.5 5.6
Mature 50+ 17 5.2
Marital status
Single 26.9 8.2
Married 18 5.4
Qualifications
None 21.7 7
‘O’ level & equivalent 25.3 7.8
Intermediate & ‘A’ level 29.4 6.4
Degree/ teaching 29.6 5.2
Job tenure
0-1 year 33.2 11.7
1-2 years 23.7 6
2-5 years 17.7 4.5
5-10 years 14.1 4
10+ years 13.4 3.6
Industry
Agriculture/energy/water/
extraction 19.8 5.4
Manufacturing 22.1 6.4
Construction 32.4 13.4
Distribution 24.8 4.5
Transport/communications 19.6 4.5
Banking 29.7 5.2
Public services 29.3 2
Private services 16.9 4.4
Source: BHPS
back into the labour market in under 6 months (about two thirds), but around 5% take longer than a year to get back and afurther 10% were still out of work at the end of the time window in the data.
Cost of job loss
Kletzer (1998) provides a useful summary of over ten years of research into the issue in North America. As yet the evidencefrom Britain is sparse, just two main papers:
• Gregory and Jukes (1997) provide the first evidence of the wage falls between jobs either side of a spell of unemploymentfor male benefit claimants (JUVOS linked with NES Panel). They find that, on average, earnings for men who have beenunemployed fall by around ten percent compared with men who remain in jobs. In addition each month of unemploymentadds about 1% to the wage fall;
• Gregg and Wadsworth (1998) use data from the British Household Panel Survey, (BHPS), over the period 1991-96 andbroaden the scope of inquiry into job displacement by including all unemployment spells (claimant or otherwise) and spellsof economic inactivity, (allowing for discouraged job seekers). They focus particularly on those suffering involuntarydisplacement. They highlight the groups which are most likely to experience displacement, those which are most likely toget back into work and the earnings changes associated with re-entry into work. They estimate the gap between old andreplacement jobs associated with involuntary job loss is 10%. Compared with those who do not lose their job it is 15%, thedifference being the foregone wage growth.
Why might job displacement involve reductions in wages? If wages rise with the experience gained within any occupation orskill group, then the job currently held is likely to pay more than any new job gained after displacement. The longer a workerhas been in the job the greater this penalty is likely to be if some or all of the returns to accumulated on-the-job experience arelost in the next job. So the costs of job loss may be higher among older and more experienced workers or wherever job loss is arelatively rare event.
This is not the only reason why we might find wage costs to displacement. Those in work seek to secure good job matchesthrough promotion or by moving firms. These matches may be provided by good firms which pay high wages and generallypromote good working conditions or just by positions that suit an individual’s talents. If workers with good jobs are unluckyenough to lose their jobs, they are unlikely to get as high quality matches on re-entry to work. For instance it has long beenestablished that certain industries offer pay premia and that unionised firms generally pay higher wages. Displacement fromthese industries, and re-employment in the non-union sector could mean a significant drop in wages.
In addition to earnings losses, new jobs are likely to be less stable. Employer or employee may decide the match is not a goodone, and the kind of redundancy rules mean that firms tend to target recent recruits when shedding labour. With variable oruncertain labour needs, firms often enshrine this practice through the use of temporary contracts. The share of full-timepermanent work in entry jobs (those taken by people saying they were out of work previously) is much lower than in theworkforce as a whole. Full-time permanent jobs make up just over 42% of entry jobs. Temporary jobs account for almost afifth of entry jobs, but just 6% of all jobs.
Tables 3 and 4 summarise the mean percentage difference in earnings between the old and new jobs for displaced workers. Asa comparison, we show the annual earnings changes recorded for those workers who remain in the same job over the year.Weekly wages of the average displaced worker are around 9% lower in the new job than in the job lost. If the worker movesfrom one full-time job to another the penalty is around 2%.1 Weekly earnings of those who remain with their employer rise byaround 5% over the year. So displaced workers not only experience wage losses relative to their previous job but they alsoforego general increases in wage levels. The total pay penalty is therefore 14% in weekly earnings, with a 9% overall paypenalty for those working full-time both before and after displacement.
There is considerable variation around these averages. Displaced women experience around twice the wage losses of men.Older workers and the least qualified also face higher monthly pay cuts than the average. The wage loss for those over 50 isaround 23%. The displaced with more than 5 years job tenure suffer losses of around 30%. Displacement tends to be rarer asjobs lengthen but the cost of losing these jobs rises greatly.
91
Table 3
Real weekly pay change by displacement status
Displaced Stayers Difference
Full-time to full-time -2.1% 7.0% -9.1%
Part-time to part-time -5.6% 2.9% -8.5%
Full-time to part-time -60.1% -53.6% -6.5%
All -8.6% 5.6% -14.2%
Source: BHPS
1 These numbers are similar to Gregory and Jukes’ (1997) findings for unemployed men. There is only a very small hourly wage penalty, on average, to beingdisplaced but this is mainly a selection effect, as the monthly wage gap is much smaller for those where hourly wages are defined.
92
Even those out of work for very short periods suffer wage losses, including the foregone wage growth the loss is around 12%.Displaced workers out of work for more than six months experience average wage losses that are twice those of people out ofwork for under one month.
The UK evidence (Gregory and Jukes 1997) suggests that about 7 percentage points of the wage gaps persist after two yearscontinuous employment. US evidence suggests that repeat job loss or broken employment generates longer persistence inlower post-displacement wages.
The levels and recent trends of pay for returnees to the labour market
We do not have data on how the cost of job displacement has changed over a long period of time, but we can say somethingabout the changing nature of wages likely to be on offer to those out of work. Table 5 gives the typical wages associated withentry jobs. These are wages earned by re-entrants into the labour market irrespective of why they left their last job. The Tablegives wages for workers in the middle of the pay distribution, (the median). Entry jobs pay wages that are considerably belowthe national average. The typical entry job pays just over £100 a week compared with around £260 for all jobs. Fifty-fivepercent of entry jobs pay below half typical earnings. In part, this is because there are relatively more part-time jobs in thestock of entry jobs than in the workforce as a whole. However, as the bottom panel of Table 5 shows, there is still a substantialwage gap in hourly wages. The typical job pays around £6.70 an hour. The typical entry job only pays around £4 an hour.Moreover the wage gap has risen over time. In the 1980s, according to the General Household Survey, entry wages fell frombeing equivalent to those received by a worker 19% of the way up the aggregate weekly pay distribution to just 16%. Morerecently the Labour Force Survey suggests a more rapid decline. Labour Force Survey data show that hourly entry wages havebeen declining in the nineties both in real terms and relative to other jobs.
Table 4
Average real weekly percent pay change by worker characteristics
Displaced Stayers Difference
Total -8.6 5.6 -14.2
Gender
Female -16.1 6.3 -22.4
Male -6.5 4.7 -11.2
Age
Youths <25 1.3 14.1 -12.8
Prime 25-49 -12.9 5.4 -18.3
Mature 50+ -22.7 1.4 -24.1
Marital status
Single -1.7 8.2 -9.9
Married -19.8 4 -23.8
Qualifications
‘O’ level & below -14 5.1 -19.1
Intermediate -10.3 5.7 -15.8
Degree/ further education 6.7 5.8 0.9
Tenure in previous job
0-1 year -6.7 18.2 -24.9
1-2 years -5.3 7.6 -12.9
2-5 years -14.4 6.2 -20.6
5+ -30 2.8 -32.8
Length of time out
0-1 month -6.6 - -
1-3 months -8.3 - -
3-6 months -13.3 - -
6+ months -13.6 - -
Source:BHPS
93
Evidence on durations of joblessness and duration dependence
The duration of unemployment has been well studied over the years but the picture of worklessness in a broader sense has beenless well documented. Evidence suggests:
• long-term unemployment rose sharply in the early 1980s but has cycled since. It has generally risen and fallen withaggregate unemployment with no clear trends;
• male inactivity has risen sharply especially for older men. Durations here are very much longer than for unemployment;
• fewer women are taking a break from the labour market around child birth and when they do it is shorter. These changes aremuted for lone parents and partners of the unemployed;
• adult members of workless households have typically not worked for 3 years or more. This has risen sharply since the early90s;
• exit rates from unemployment fall sharply with duration up to about two years. The statistical evidence is that the generaldecline in exit rates since the 1970s is not largely due to variation in who becomes unemployed. But the long standingpattern of declining exit rates with duration has never fully been established as being due to individual or area heterogeneityor true duration dependence;
• an alternative approach is to look at the impact of schemes that get the long-term unemployed into work ortraining/education. Random assignment experiments in US suggest that work and work based training are useful in raisingemployment (and wages) especially for older people and lone parents. Raising job search and job search skills areespecially helpful. This suggests that there is al least some duration dependence and heterogeneity in search effectiveness.
Long-term unemployment (12+ months) accounted for around 20% of the unemployed in the 1970s. After the 1980s recessionit has cycled from 30 to 40% (see figure 1 and 2 taken from Nickel, 1998). There appears to be a broadly stable relationshipbetween the level of unemployment and the share that is long-term unemployment; (Machin and Manning (1998)).
It is important to remember that the unemployed make up a fraction of those not in work (around 20%) and among the inactive
Table 5 The changing pattern of real wages in entry jobs
Median Entry wage in overall distribution(%) (%)
Weekly wages
GHS
1980 104.2 19.1
1984 107.3 19.3
1988 106.7 17.6
1990 107.1 15.7
LFS
1993 99.8 17
1995 93.9 15.3
1997 99.2 16
Hourly wages
GHS
1980 3.63 22.3
1984 3.73 20.8
1988 3.98 19.2
1990 3.98 16.6
LFS
1993 4.35 23.2
1995 4.17 20.7
1997 4.02 19.3
Source:LFS 1993-1997, GHS 1980-1990. GHS data are 3 year averages centered on year shown. Figures in April 1998 pounds.
94
patterns of durations out of work have changed considerably. The inactive have switched sharply from primarily being womenwith young children and a working spouse to those who are members of workless households. In 1970s about 20% of inactivewere in workless households. This now stands close to 60% (see Gregg, Hansen and Wadsworth, 1998). Older men especiallyhave seen sharp rises in inactivity (Campbell, 1999) whilst lone parents and partners of the unemployed have not increasedattachment to the labour market in the way other mothers have. Of adults in workless households some 60% have not workedin at least the last three years. This has risen sharply, up from 45% in the early 1990s. The new DSS benefit data base alsoshows that 56% of those claiming benefits while out of work have durations over 2 years in length.
Duration dependence
The flash card evidence for duration dependence is the rapid decline in outflow probabilities with elapsed duration. Evidencedeveloped by Layard, Nickell and co-authors suggests, plausibly, that the general decline in out flow rates after the 1970scould not be attributed to individual characteristics (see Layard, Nickel and Jackman, 1991, for a good summary). But howmuch of the long standing relationship between duration and exit rates is down to duration dependence is unclear. Estimationand controls for unobserved heterogeneity rely on assumptions which may or may not be believed. However, whilst in UK datastatistical studies usually find strong evidence of duration dependence, it is not found for similar data on the continent (seeMachin and Manning, 1998). This tends to suggest that it is not the particular techniques used that drive the evidence ofduration dependence in the UK.
Another approach is to look at the evidence that schemes to re-attach the long-term unemployed to work are successful. USrandom assignment experiments net out unobserved heterogeneity. A S Krueger et al (1998) treasury paper suggests that atleast for older displaced workers and lone parents job subsidies modestly raise future employment rates (and wages) aftersubsidy has stopped. This does imply that to some extent time spent in work (or to a lesser extent make work schemes) raisesthe individuals chances of securing a job match.
The evidence on concentration of unemployment over long periods and repeat spells
The duration of spells of unemployment highlights that most people leave unemployment quickly but most of unemployment isconcentrated on a smaller group. Here it is interesting to look at how repeat incidence could change this picture. There is not alarge body of existing work but Machin and Manning (1998) present data from BHPS and comparisons with US and Germandata. Teasdale (1998) presents summary statistics from the JUVOS panel of the claimant unemployed:
• about 30% of the working age population has a spell of unemployment in a five year window. About half of these peoplehave more than one claim. It is higher for men than women and has the highest incidence for people in their twenties;
• Machin and Manning suggest 78% of unemployment in a year falls on those who are unemployed for 6 months or more.Over a four year window about half falls on those with 2 or more years unemployed (not necessarily in one spell);
• these figures are akin to Germany but way above the US where unemployment is spread more evenly;
• they do the same for non-employment (including the inactive). Here, over one year, nearly 90% of time spent out of work ina year falls on those out for 6+ months and 70% of worklessness over four years falls on those out for 2+ years;
• about 45% of those signing on have terminated a previous claim in the last 6 months. Calculations based on mid-points ofbands of Tables in Teasdale suggest that 49% of all unemployment in a five year window fell on those with 2 plus years upto 1987 and this was up to 55% in the five years up to 1996. But this could be cyclical.
Is the evidence that unemployment now increases the risk of unemployment causal?
There are three possible reasons why unemployment now is associated with unemployment in the future. That some people arealways more prone to unemployment because of low education etc. That high areas of unemployment persist (although theregional pattern has varied quite a lot over the last thirty years). Or finally that being unemployed hurts your futureemployment prospects (or structural dependence). Sorting out these effects has all the same problems as for durationdependence above.
The few studies in this area for the UK suggest positive structural dependence over and above duration dependence discussedabove and that such effects are more marked after 25 or so (Narandranathan and Elias, 1993, Arulampalam and Booth, 1997,McCulloch and Dex, 1996). The National Child Development Survey is a particularly powerful tool in getting at this issue. Byfollowing a cohort of individuals all born in the same week and following them through childhood into adulthood, we canobserve much more about the individual than is normal in labour market studies. There is also no aggregate macroeconomiccycle to worry about as they are all the same age when any shocks occur. Furthermore we can observe those that move areaand use this to separate effects of being in a bad local labour market from individual experiences. The NCDS asks details ofindividuals experiences in the labour market from 16 to 23 when they were aged 23 and from 23 to 33 at age 33. Table 6 showsthe individual patterns across these two spells. It gives the proportion of time spent unemployed or non-employed between 23and 33 broken done by how many months they were unemployed between 16 and 23. The results are striking. It shows that forthose men who had no unemployment before 23, only 1.6% of months between 23 and 33 were spent unemployed. For thosewith some unemployment but less than 6 months, this rises to 4% of months and for those with more than a yearsunemployment, an impressive 23% of months between 23 and 33 (or about 2 years). The patterns for women are less acute but
95
are stronger for non-employment which includes looking after children.Table 7 controls for the characteristics that might lie behind such correlations. Column one reports the raw differences impliedby a Tobit estimation as compared to a base group of those who had no unemployment before 23. Column 2 controls for regionof residence in at age 33. Column 3 adds educational qualifications and column 4 adds all parental background, ability testscores behavioural tests etc. that NCDS undertakes on these children. It shows that little of this correlation can be explained bythese factors. They reduce the implied impact of a years extra unemployment before 23 on the amount of time spentunemployed from 23 to 33 from adding 21% to 18%. Education has the biggest effect. This is not because education etc are notstrongly correlated with a persons chances of being unemployed: they are. It says that for those who do have a lot of earlyunemployment, of which only a minority are well educated, they have a lot more later unemployment whether they are welleducated or not. Table 3.2 looks at men only for brevity but patterns are similar for women. Further extensions to focus onmigrants from high unemployment regions to low unemployment regions sees correlations which are only a fraction less strongas those for all people. Taken together with more sophisticated econometric work, not reported but see Gregg (1998), suggeststhat about two thirds of the raw correlation between higher early unemployment and higher late unemployment is down to theunemployment itself and one third because of area or individual variation. This is a strong justification for early intervention
Table 6: The Effects of Youth Unemployment: New Results from the National Child Development Survey Sample Means
Notes: 1 Tobit estimates; number of observations = 3903; t-ratios in brackets.2 Mean of the dependent variable in regressions (1) to (4) is 4.2% (i.e. on average, men in the sample spent 4.2% of their time unemployed between 23-33).3 Mean of the dependent variable in regressions (5) to (8) is 5.4% (i.e. on average, men in the sample spent 5.4% of their time inactive between 23-33).
Inactivity here means time not spent in employment or education. It therefore includes time spent unemployed, sick, and at home. 4 The variables denoting spells of unemployment between 16 and 23 are 0/1 dummies (i.e. they take the value 1 if the individual spent the specified period in
unemployment between the ages of 16 and 23).5 Eleven regional dummies are included to pick up regional effects.6 Eight education dummies, representing the highest qualification achieved by the individual, are included to proxy for the effects of education on the
probability of unemployment.7 The variables include individual characteristics like ability aged 7, behaviour at 7 and 16, absenteeism, special needs, suffered major/minor illness,
involvement with the law, along with parental education, parental unemployment, lone parenthood, financial measures and ethnicity.
Table 7: The Effects of Youth Unemployment: New Results from the National Child Development Survey (A) Men
96
Table 8: The Effects of Youth Unemployment: New Results from the National Child Development Survey (B) Women
Notes: 1 Tobit estimates; number of observations = 4233; t-ratios in brackets.2 Mean of the dependent variable in regressions (1) to (4) is 2.4% (i.e. on average, women in the sample spent 2.4% of their time unemployed between 23-33).3 Mean of the dependent variable in regressions (5) to (8) is 24.2% (i.e. on average, women in the sample spent 24.2% of their time inactive between 23-33).
Inactivity here means time not spent in employment or education. It therefore includes time spent unemployed, sick, and at home. 4 The variables denoting spells of unemployment between 16 and 23 are 0/1 dummies (i.e. they take the value 1 if the individual spent the specified period in
unemployment between the ages of 16 and 23).5 Eleven regional dummies are included to pick up regional effects.6 Eight education dummies, representing the highest qualification achieved by the individual, are included to proxy for the effects of education on the
probability of unemployment.7 The variables include individual characteristics like ability aged 7, behaviour at 7 and 16, absenteeism, special needs, suffered major/minor illness,
involvement with the law, along with parental education, parental unemployment, lone parenthood, financial measures and ethnicity.
for youth to prevent long-term unemployment, if such interventions work. ReferencesArulampalam, W. Booth, A. and Taylor, M. (1997) Unemployment Persistence mimeo Department of Economics WarwickGregg, P. (1998) The Scarring Effects of Youth Unemployment: New results from the NCDS, mimeo, CEP, LSEGregg, P. Hansen, K. and Wadsworth, J., (1998), The Rise of the Workless Household in Gregg and Wadsworth (eds) The State of Working Britain, MUP,Manchester forthcomingGregg, P. and Wadsworth, J., (1996), ‘Mind the Gap: The Changing Distribution of Entry Wages in Great Britain’, Centre for Economic PerformanceDiscussion paper no. 303.Gregory, M. and Jukes, R., (1997), ‘The Effects of Unemployment on Subsequent Earnings: A study of British Men, 1984-94’, Department of Education andEmployment Working Paper. Huff Stevens (199*)Kletzer, L., (1998), ‘Job Displacement’, Journal of Economic Perspectives, Winter, pp. 115-36.Kreuger, A. (1998) treasury paperLayard, R. Nickell, S. and Jackman (1991) Unemployment: Macroeconomic Performance and the Labour Market, OUP, OxfordMachin, S. and Manning, A. (1998) The Causes and Consequences of Long-term Unemployment in Europe Centre for Economic Performance LSE DiscussionPaper no. 400McCulloch, and Dex, S. (1996) Modelling Male Unemployment Persistence in Britain mimeo University of CambridgeNaradrathan, W. and Elias, P. (1993) Influences of Past History on the Incidence of Youth Unemployment: Empirical Findings for the UK, Oxford Bulletin ofEconomics and Statistics Vol. 55 No. 2 pp 161-185Nickel, S. (1999) Unemployment in Britain in Gregg and Wadsworth (eds) The State of Working Britain, MUP, Machester forthcomingOECD (July 1996) ‘Making work pay’ Chapter 2, pp25-58, Employment Outlook 1997. Organisation for Economic Co-operation and Development, Paris.Teasdale, P. (1998) The Incidence of Unemployment and Repeat Spells Using Claimant Unemployment Data in Labour Market Trends, November, HMSO.
97
Session 4: Work and Poverty: Discussion
In open discussion on the presentations, the following points were made:
Views on work as an anti-poverty strategy
Current Government policy on poverty focuses heavily on encouraging work, viewing the lack of work as the main cause ofpoverty. This differs from past policy, to a degree, which provided more help for the poor through the welfare system. It is notclear that the Government’s labour market policy (involving re-engagement to the labour market, reforms to make work payand investment in skills and education) is also a comprehensive anti-poverty strategy.
Initial engagement to the labour market is seen as crucially important. For example, Paul Gregg’s work suggestsunemployment spells early in a person’s life can have detrimental effects. However, initial engagement may not be sufficient toalleviate poverty in the longer term. Indeed it may actually be counter-productive in terms of future earnings and later labourmarket success if initial engagement is to a low-paid job. An implication of the work on the low-pay, no-pay cycle is that notall jobs are good jobs.
Given the existence of the low-pay, no-pay cycle it was suggested that the Government may need to go further than initialengagement policies, placing more emphasis on how we develop people from their first job. Such increased support wouldaim to ensure individuals remain employed in the longer term and move up the earnings ladder. This may be achieved bycounselling people (which is occurring to an extent in the New Deal, with an on-going case management approach) or morefundamentally by changing the nature of the low wage labour market itself, for example to reduce turnover in low wage jobs.
One of the best safeguards against low pay is having been with your employer for a long time. Very few of the low paid havebeen with their employer for more than 1 or 2 years. This doesn’t mean that we should keep these jobs going, however, asextending the life of low paid jobs doesn’t mean they will become good jobs; it is simply the nature of many low paid jobs, thatthey last only a short time.
Possible explanations for the existence of the low-pay, no-pay cycle included:
• low paid jobs may signal low productivity to employers;
• low paid jobs may deskill, and perhaps more importantly do not skill people, reducing or failing to enhance worker’sproductivity;
• working in a low-paid job may restrict the amount of effort an individual can devote to search for alternative employment.
Further research should look into which of these, or other explanations for the low-pay, no-pay cycle, are of most importance.
Education was felt to help in the labour market by preventing individuals from becoming either low paid or unemployed.However, education doesn’t help individuals get out of these labour market states as much as one might expect. In effecteducation has a strong effect on the level of earnings, but a lesser effect on changes in earnings.
The ability to predict those susceptible to the low-pay, no-pay cycle would, obviously, aid understanding. However, bydefinition with scarring one cannot perfectly predict those who will be subject to the low-pay, no-pay cycle. Low pay, of itself,scarsan individual (whoever they are) to the low-pay,no-pay cycle. Past labour market histories can, to an extent, be used topredict – as those in the low-pay, no-pay cycle will have been through it before.
The existence of the low-pay, no-pay cycle was felt to have interesting implications for early intervention. It was felt thatshocks and events which occur throughout an individual’s life may require ‘recovery’ action. For instance, no form of earlyidentification would have helped, now redundant, coal-miners. In essence there may be macroeconomic and structural changeswhich need to be adapted to.
Individual, not just labour market, characteristics need consideration
It is often the case that those with labour market problems simultaneously suffer multiple problems. For example, some may havechildren but no access to childcare or they may be long-term sick/disabled. Therefore, it may be an individual’s personalcharacteristics, rather than their labour market characteristics, that cause them problems in holding down a job for any period oftime. When providing help, policies need to consider personal characteristics and circumstances as well as labour market factors.
Individuals may not only be scarred in terms of their future labour market experience by unemployment. There is also someevidence of a relationship between unemployment and health. A study, using the NCDS, found in 33 year old men that ahistory of unemployment increased their chances of ill-health. In contrast it seems that stable income perpetuates good health.Work by Andrew Oswald reinforces this point. He has found that happiness (if one accepts that we can measure happiness)falls with unemployment and it falls further with the duration of unemployment. Chillingly, suicide rates also increase withunemployment and increase further with duration of unemployment.
97
98
Changing the balance of demand and supply of labour
Increasing the supply of highly skilled people and the demand for low skilled people was seen as a possible way of narrowingthe earnings distribution. However, whilst changing the relative balance of demand and supply would change wages at thebottom end of the labour market, there will always be a bottom end. It was felt that there may be a particular problem withpeople who end up at the bottom of the labour market, especially in terms of lifetime equality. Therefore whilst changing thedemand and supply of labour will change the returns to education (through changes in wages) it is questionable whether thiswill be enough to deal with the scars caused by the experience of unemployment or low pay. In addition increasing the supplyof highly skilled workers, or moving people up the skills distribution, may be problematic given the repeatedly poor findings ofgovernment training programs.
The problem is simply a lack of demand
It was pointed out that the generation which entered the labour market in the 1930s had a huge experience of unemploymentand yet in the post-war ‘golden age’ period, attachment to the labour market, in terms of the participation rate, was at itshighest ever. This would seem to suggest that unemployment did not detrimentally scar these people. It may be that with 2million people unemployed and large numbers inactive the main problem is simply deficient labour demand. Indeed the labourmarket may be working extremely well given a general shortage of jobs. It was felt there is a shortage of skilled jobs and thatthe economy needs to be able to create more skilled jobs.
A large part of the problem was seen as being due to withdrawal from the labour market, rather than specificallyunemployment. For instance, if there had been a pick up in economic activity in the labour market in this recovery interestrates would not have needed to increase and so growth could have increased by more. Therefore the fact that some sections ofthe population are withdrawing from the labour market may be exacerbating demand problems, by inhibiting the economyfrom growing too quickly.
Session 5: Childhood Poverty and Family Structure
‘Poverty in the early years: evidence from the 1958 and 1970 British Birth Cohort Studies’
John Bynner, Centre for Longitudinal Studies, Institute of Education
Introduction
Children’s poverty is imposed rather than acquired through experience. Yet at the same time, growing up in poverty is a keyfactor in determining what adult experience, especially in the labour market, is likely to be. In this respect, childhood povertyplaces limitations on the individual child’s development, relative to others, as expressed in the idea of “deprivation”. In hisreview of the effects on children of poverty and inequality in the UK, Kumar (1996), for example, refers to “the lack ofadequate resources to satisfy certain essential minimum human needs”. But he goes on to say that this apparently “absolutistdefinition” still has to determine where the minimum standards are set, reflecting a degree of relativity in the term. Even Sen’s(1978) identification of poverty as with “an irreducible core of deprivation” is still relativistic in the sense that deprivationembraces not only lack of resources to meet basic physical needs, but those needs which are more likely to be culturallyacquired. Families living in homes without bathrooms or even telephones exhibit a degree of deprivation today that would nothave been recognised sixty years ago.
Thus poverty is both defined by circumstances and, to a certain extent, determines how those circumstances affect life. Indiscussing “Children at Risk”, OECD (1996) adopt a definition of “risk” from Schoor (1988), in which “non school factors”affecting school failure and adolescent problems, including delinquency, comprise:
• growing up in persistent or concentrated poverty, and in a family of low social class;
• being born unwanted or into a family with too many children born too close together;
• growing up with a parent who is unemployed, a teenager, a school dropout or illiterate;
• having a parent who is impaired as a result of alcoholism, drug addiction or mental illness and/or a parent who is withoutsocial support;
• growing up in a family or neighbourhood with such a high level of social disorganisation as to leave a young childunprotected from abuse and violence, and with little exposure to healthy role models;
• growing up outside ones family, especially in multiple foster care or institutional placements;
• growing up with the sense that one has bleak prospects for good employment or a stable family life and little power toaffect one’s own destiny and that one is not valued by the outside world.
The centrality of persistent poverty in this list emphasises its significance in lying at the core of disadvantage. This also makesthe point that it is in combination with other factors that the potency of poverty for damage to life chances is most evident. Achild may be able to withstand any one or two of such “risk” factors, but in combination they constitute severe barriers tonormal and healthy development. Endean and Harris (1998) (from DSS and ONS respectively) extend this idea to thedefinition of poverty itself, embracing among their “non-income indicators”, child health, education and qualifications,workless households, multiple deprivation. The DSS area-based index of deprivation similarly ranges beyond the confines offamily income, including among its twelve indicators: unemployment, receipt of benefits, mortality rates, low educationalattainment, derelict land, crime availability of basic household amenities, and household overcrowding.
Research resources
The study of the consequences of early childhood poverty in the shorter and longer term for people’s life chances is greatlyaided by the availability of the large-scale longitudinal birth cohort studies that are unique to Britain. Each has involvedfollowing up one week’s cohort of babies from birth into adult life, starting in 1946, 1958 and 1970 respectively. All havecontributed to understanding the adverse effects of childhood poverty. For example James Douglas’s pioneering studies of theeffects of social class on school achievement, such as The Home and the School (Douglas, 1964) were based on the 1946study. More recent studies such as Kuh and Wadworth’s (1991) analysis of the effect childhood influence on adult earningshave similarly used this source. However, the birth cohort study that has been particularly concerned with the social outcomesin adulthood of earlier child circumstances is the 1958 birth cohort or “National Child Development Study” (NCDS) (Ferri,1993). Less widely used or well known, but in many respects particularly relevant to current concerns is the 1970 British BirthCohort Study (BCS70) (Bynner, Ferri and Shepherd, 1997). In this paper I want to draw on examples of the use of these twolatter longitudinal data sources to examine three questions:
• How is child poverty measured?
• What are the short-term, medium term and long-term effects of child poverty on later life chances?
• Are the effects of child poverty changing in successive cohorts?
99
How is child poverty measured?
A variety of approaches to the measurement of poverty, have been adopted by different researchers working with the birthcohort studies. These are informative in revealing different theoretical and methodological predilections and also help toexplain some of the inconsistencies in particular research results; though overall coherence in the patterns of relationships isevident. Definitions vary from the most narrow, concerned with income-based definitions of poverty, (e.g., Gregg et al, 1998),to those that extend to material circumstances (e.g.Wedge and Prosser, 1973). The broadest definitions extend to culturalresources, including the parents’ level of education (e.g. Osborne and Milbank, 1987). At this level, poverty or deprivation isnow frequently translated into an even broader generic notion that of social exclusion, i.e. economic and cultural deprivationare obstacles to the child’s prospects, which in a sense are seen as the responsibility of the state. Figure 1 summarises the rangeof approaches.
Figure 1 Poverty Definitions
Narrow definition Broad definition Broadest definitions
Low family income Poverty Deprivation Social Exclusion
economic economic economic disadvantageindicators material material in antecedents
indicators cultural disadvantageindicators in outcomes/
statuses
Associated with and, in a sense, underpinning all of these definitions is the ubiquitous social class based on the RegistrarGeneral’s classification of (usually) father’s occupation. In the birth cohort studies, the first social class measures were thosederived at birth, and subsequently those taken in the later sweeps of the studies: at ages 7, 11, 16, 23 and 33 in the case ofNCDS, and 5, 10, 16 and 26 in the case of BCS70. Through adulthood (post 18), of course, the cohort member’s social class asdefined by their own occupation, comes into the picture alongside that of their parents.
Beyond the questions of concept coverage researchers also differ with respect to the assumptions of their measurement method.For example, some researchers take the absolute level of a poverty indicator, such as family income, or use it to identifyrelative magnitude, e.g., bottom quartile range. Another variant resides in whether cross-sectional poverty states are used inlongitudinal models as opposed to cumulative measures across states at different times e.g. persistent as opposed to transientpoverty. Some of the more complex indices of poverty embrace cause and effect within the measure itself. For example,parents’ education in some formulations is treated as exogenous to parents’ occupation on which social class is based. In othersboth are treated as exogenous.
Some examples of the different measurement approaches are set out below.
• Wedge and Prosser: Born to Fail (1973):Using NCDS up to age 11, Wedge and Prosser defined poverty in terms of threebroad sets of indicators: family size – one parent family or large family (23%); income – child receives free school meals orfamily on supplementary benefit (14%); housing – more than 1.5 people per room, or where there was no hot water supplyfor the family’s exclusive use (23%). Their poverty index comprised counts of the number of characteristics present in thechild’s family. On this basis the most disadvantaged group of all, exhibiting all characteristics, comprised 6.2% of theNCDS sample. This contrasted with the group who showed none of these characteristics 63.9%.
• Osborn and Milbank: The Effects of Early Education (1987):Using BCS70 data up to age 10 Osborn and Milbanks“social index” is a much broader concept of deprivation than Wedge and Prosser’s, embracing cultural as well as economiccomponents. It comprises 7 indicators: social class of father, highest educational qualification of either parent, housingtenure, type of accommodation, persons per room ratio, car ownership and telephone availability. Those exhibiting the mostpoverty, i.e. who showed the poverty attribute on all these indicators, comprised between 6% and 8% of the whole sample.
• Hobcraft: Intergenerational and Life Course Transmission of Social Exclusion (1998).Using the complete NCDSdataset to predict adult outcomes up to age 33, Hobcraft constructed a longitudinal index based on just two indicators:reported financial difficulties at 7, 11 and 16, and free school meals at 11 and 16. The highest levels of poverty would beshown by existence of these attributes in the cohort members’ families at all ages and stages. The most affluent familieswould show none of them.
What are the short-term, medium term and long-term effects of child poverty on later life chances?
Wedge and Prosser’s book, Born to Fail, was one of the first to demonstrate, using NCDS, the strong effects of early childhoodpoverty on subsequent performance through primary school. Wedge and Prosser calculated that at age 11 disadvantaged childrenwere, on average, 3.5 years behind the other children in reading and maths test scores, though one in seven actually did betterthan half the other children. They noted that the key predictor was social class, concluding that the ethos of the British primaryschool was antipathetic to the educational progress of their largely working class intakes. Doria Pillings’ book Escape fromDisadvantage (1990) reported a follow-up by interview at age 27 of those individuals who as children had been defined byWedge and Prosser as disadvantaged, yet subsequently had succeeded in education and in the labour market. She found that a
100
common feature of their experience was sustained parental aspirations and commitment from teachers, which overrode theeffects of what was often temporary hardship. In other words childhood disadvantage is both pervasive and permeable in itseffects. It inhibits life chances, but given a particular family and school context, it need not be permanently disabling.
Osborn and Milbank’s study focused on a particular feature of the way middle class parents build the foundations for theirchildren’s school achievement – the provision of pre-school educational experience. They found that such experience,especially the pre-school play group, was significantly more common in families that scored high on their social index(absence of deprivation). They also found in a multivariate analysis that although pre-school experience (excluding the LocalAuthority day nursery) and social index score predicted reading and maths performance at age 10, the social index score wasmuch the stronger predictor. Of the various kinds of pre-school provision, the largely middle class home-based playgroup withsmall numbers of children again came first.
When does the social gradient in educational achievement first become evident? Using the 1970 cohort we can go back tocognitive development before the age of 5, taking advantage of two early data sets based on sub-samples. Feinstein’s (1998)analysis of BCS70 data collected on cognitive development at 22 months and 42 months, demonstrates a social class gradientin cognitive development indicators beginning at 22 months, which expands through 42 months and up to age 5. In otherwords, even in early infancy, not only is a social gradient apparent but this is accentuated with every subsequent month of life.
Modelling poverty outcomes in childhood and adulthood
The studies reviewed so far demonstrate strong social effects on children’s subsequent progress. The studies by Gregg et al(1998) and Hobcraft (1998) using NCDS data (both reported in this volume) extend the findings to adulthood, demonstratingthe persistent effects of childhood poverty, over and above other variables, on a number of adverse (social exclusion) outcomesin adulthood including low income. Such findings parallel those that are reported in more detail below on the origins of basicskills deficits in adulthood (Bynner and Steedman, 1995; Parsons and Bynner, 1998.)
Are the effects of child poverty changing in successive cohorts?
From the research reviewed above the connection between childhood poverty and adult disadvantage is well established. Themediating factor is educational achievement on which persistent poverty exercises an inhibiting effect. The last issue toconsider is whether the relationship between poverty/class and education and their adult outcomes is changing over time. Isthere a cohort effect? To find out NCDS and BCS70 data were used to model separately for each cohort the effects ofchildhood poverty on two kinds of outcomes: educational attainment in childhood and experience of unemployment inadulthood (Bynner, 1998). Using representative 10% samples of each cohort in which adult literacy and numeracy weremeasured (Bynner and Steedman, 1995; Parsons and Bynner, 1998), the longitudinal data were used to encompass the amountof unemployment experienced since leaving school up to age 21 (BCS70) and up to age 23 (NCDS).
Three childhood poverty indicators were employed: “free school meals”, “overcrowding in the home” and “father’s socialclass”. Notably there were changes sometimes in opposite directions in the prevalence of these characteristics in the two birthcohorts. In the case of free school meals at age 11, NCDS, and age at 10, BCS70, there was a rise from NCDS to BCS70 –10% to 16%. In the case of overcrowding, there was a marked decrease (65% to 30%). The proportion of fathers inprofessional occupations increased between the two cohorts (20% to 27%) and the percentage in unskilled manual jobsdeclined (22% to 18%). In other words, there was increasing disadvantage associated with free school meals, whiledisadvantage associated with overcrowding and social class was reducing.
To identify the short and longer-term effects of family poverty within a comprehensive model, structural equation modellingusing the LISREL program was applied. The model specified adult unemployment related to a range of variables measuredover the period 16 to 21/23. These variables were themselves related to variables measured at 10 and then before 10 back tobirth. The effectiveness of the model was assessed in terms of estimated multiple correlations squared. These signified thepercentages of variance in outcome variables at different ages that could be accounted for in terms of all the variables includedin the model measured prior to them. The program also gave the standardized regression coefficients, or ‘path’ coefficients,between the variables measured at different time-points, indicating the strength of each of the longitudinal relationships in themodel, holding constant the effects of all other variables. Variables were included only if they had been measured in the sameway in both BCS70 and NCDS (Table 1). Full details are given in Bynner (1997).
Table 1: Adult outcomes of childhood poverty: variables included in the model
• At 21 (BCS70) and at 23 (NCDS): amount of unemployment experienced since leaving school;
• 16 up to 21 (BCS70) and 16 up to 23 (NCDS): school leaving examinations score at 16, adult literacy score, adultnumeracy score, age left full-time education, number of jobs, malaise (measure of depressive mood), whether experiencedany work based training;
• At 10 (BCS70) and at 11 (NCDS): maths test score, reading test score, mother’s interest in education, father’s interest ineducation, overcrowding, rented housing;
• At birth: age mother left full-time education, age father left full-time education, family social class (assessed from father’soccupation).
101
There were two notable results from the analysis of particular relevance to our interest here. First, in relation to the explanationof adult unemployment outcomes, there were marked differences between the cohorts. 16+ attributes and experiences played alarger part in explaining unemployment in the 1970 cohort than in the 1958 cohort as reflected in the percentages of varianceexplained -11% in the case of BCS70 and 5% for NCDS. This suggested that personal attributes were more at a premium inemployability for the 1970 cohort than for the 1958 cohort for whom unemployment was much more labour market dependent.Those attributes to do with education (exam score, numeracy and literacy), as we know, are rooted in childhood experience, sowe can conclude that the effects of childhood poverty on the experience of unemployment was stronger in the younger cohort.
The second result of interest is the relationships between variables at the other end of the model, i.e. those relating theantecedent conditions at birth to attributes at age 10 (BCS70) and at age 11 (NCDS). In the 1958 cohort there were muchstronger connections between the two, especially between social class and the 10-year (BCS70) and 11-year (NCDS)outcomes. Maths and reading scores, mothers interest in education, fathers interest in education, overcrowding and rentedhousing were all predicted by family social class at birth in the 1958 cohort. In contrast, in the 1970 cohort, the relationshipswere much weaker or non-existent. For example, there was no connection between family social class at birth and the motheror father’s interest in education, and there were weaker relationships with maths and reading. For the 1970 cohort these interestvariables were more strongly related to the age the mother had left full-time education. Only overcrowding and housing tenurewere related to social class in the 1970 cohort but more weakly than in the 1958 cohort.
These results point to a greater degree of separation of class-based origins from subsequent experiences and achievements inthe younger cohort – a greater degree of fluidity in life chances than was the case for the older NCDS cohort.
Conclusion
This paper started with a consideration of the various meanings of childhood poverty, and the different approaches tomeasuring it in the 1958 and 1970 British birth cohort studies. It ended with models of the effects of childhood poverty onadult life chances. The variety of approaches presents problems in drawing firm conclusions about whether it is poverty,deprivation, general disadvantage or social class that lies behind educational difficulties and subsequently adult socialexclusion. Indexes constructed from a number of indicators simplify analysis but obscure some of the nuances in relationshipsthat may be particularly significant for policy. For this reason single indicator variables which can be mapped across studiesare for most purposes preferred.
Overall though the patterns of relationship are consistent in pointing repeatedly to the damaging effects of childhood povertyon later educational achievement and adult employment prospects and that such poverty is strongly associated with socialclass. The weakening of the class effect in the younger (BCS70) cohorts points to a possible loosening of old boundaries andweakening of structural imperatives which can be seen as encouraging for policy interventions. Whether this trend will persistremains an open question for which further research, e.g. on cohort members’ children, is needed to resolve. In the meantimewe can conclude that resources directed at families and children, including the provision of educational opportunities for theparents, are likely to break more effectively into the vicious circle of class, poverty and educational disadvantage than was thecase in the past.
References
Bynner, J. (1998) ‘Education and Family components in the transition from school to work’, International Journal of Behavioural Development,22, 29-53.Bynner, J. and Steedman, J. (1995) Difficulties with Basic Skills. London: Basic Skills Agency.Bynner, J , Ferri, E. and Shepherd, P. (eds.)(1997) Twenty-Something in the 1990s: Getting on Getting By, Getting Nowhere.Aldershot: Ashgate.Douglas, J. W.B. (1964) The Home and the School,London MacGibbon and KeeEndean,R. and Harris, T. (1998) ‘A Picture of Poverty in the UK’, Social Trends Quarterly Pilot edition,Office of National StatisticsFeinstein, L.(1998) Pre-school Educational Inequality? British children in the 1970 cohort, Discussion Paper 404, Centre for Economic Performance, LondonSchool of Economics.Ferri, E. (ed.) (1993) Life at 33.London: National Children’s Bureau.Gregg, P., Harkness, Machin, S. and Thomas, J. (1998) Child Development and Family Income Report for the Joseph Rowntree Foundation,mimeo.Hobcraft, J. (1998) Intergenerational and Life Course Transmission of Social Exclusion: Influences of Childhood Poverty, Family Disruption and Contact withthe Police, Case-paper 15, Centre for the Analysis of social Exclusion, London School of Economics.Kuh, D. and Wadsworth, M. (1991) ‘Childhood Influences on Adult Male Earnings, British Journal of Sociology,42, 537-55.Kumar, V. (1993) Poverty and Inequality in the UK: The Effects on Children,London: The National Children’s Bureau.Organisation for Economic Co-operation and Development (1996) Our Children at Risk,Centre for Educational Research and Innovation, Paris: OECD.Osborn. A.F. and Milbank, LE. (1987) The Effects of Early Education.Oxford: Clarendon Press. Parsons, S. and Bynner, J. (1998) Influences on Adult BasicSkills. London: Basic Skills Agency.Pilling, D. (1990) Escape from Disadvantage.Basingstoke: the Falmer Press.Sen, A. (1978) Three Notes on the Concept of Poverty,Geneva: International Labour Organisation.Wedge, P. and Prosser, H. (1973) Born to Fail. London: Arrow Books.
102
103
‘Early childhood interventions and outcomes’
Jane Waldfogel, Columbia University and Centre for Analysis of Social Exclusion, London School of Economics
Introduction
Recent advances in brain research have provided new evidence that experience in the earliest days, weeks, and years of lifematters. The human brain, we now know, grows very rapidly in the first three to five years of life, and what happens in thosefirst years can either promote development or curtail it.1
This new evidence from brain research has greatly increased interest in the effects of early childhood interventions onoutcomes for children. In this paper, I lay out some ground rules for the analysis of early childhood interventions andoutcomes. Then I consider what we know about potential benefits and ill effects, and then conclude with comments about whatwe don’t know.
Ground rules for the analysis of early childhood interventions and outcomes
Before reviewing the evidence on early childhood interventions and outcomes, it is important to establish some ground rulesfor the analysis.
The first is that one must be clear about what type of intervention one is analysing. Early childhood intervention and childcareare not synonymous. Early childhood intervention refers to programs such as childcare or home visiting that are designed topromote the development of children from birth through the time they enter school, and typically these programs are targetedto children identified as high-risk for poor development. Childcare, in contrast, is not always designed primarily as an earlychildhood intervention, and may be targeted to other groups (for instance, the children of employees or students). Childcare isvery heterogeneous, with provision ranging from childminders, babysitters, and nannies to playgroups and nurseries and pre-schools.2 Moreover, we do not know very much about the quality of childcare being offered in most childcare settings.3 Yet,we know that quality of childcare matters for child outcomes.4 Thus, in reviewing any study of early childhood interventionand outcomes, it is important to establish what model of intervention was provided, whether it included childcare, and, if so,what, if anything, we know about the quality of that care. It is also important to think about what the intervention was meant toprovide; some models, for instance, place more weight on cognitive development than others.
The second ground rule is that one must be clear about when the intervention was provided. In the case of childcare, there is agreat deal of evidence that childcare begun in the first year of life has a different effect on later emotional adjustment than carebegun thereafter (Haskins, 1985; Belsky and Eggebeen, 1991; Baydar and Brooks-Gunn, 1991; Smith, 1994; Bates et al, 1994).The same may be true of cognitive development, with childcare begun in the first year of life appearing to have negativeeffects for some groups (Desai, Chase-Lansdale, and Michael, 1988; Blau and Grossberg, 1990; Baydar and Brooks-Gunn,1991; Smith, 1994), while care after the first year of life seems to have positive effects (Blau and Grossberg, 1990; Baydar andBrooks-Gunn, 1991; Brooks-Gunn, Liaw, and Klebanov, 1992; Brooks-Gunn et al, 1993).5 The few studies that have been ableto control for childcare quality find that it plays an important mediating role (Vandell, Henderson, and Wilson, 1988; Field,1991; NICHD, 1997), as does the type of care (Howes, 1988 and 1990; Baydar and Brooks-Gunn, 1991; Field, 1991; Smith,1994). It may also matter whether the care was full-time or part-time.
The third caution is that one must be clear about which children received the intervention. Again using childcare as anexample, the age at which a child enters childcare is obviously a critical mediating factor, but so too are factors such as thechild’s attributes, family background, and current living situation. These characteristics may influence both the type ofchildcare used and the child’s outcomes; thus, if child and family characteristics are not properly controlled, one mayerroneously attribute outcomes as the result of childcare when they are in fact the result of other factors. Further complicatingthe analysis is the fact that childcare and family characteristics may have an interactive effect. For instance, many studies havefound that children from families that are economically disadvantaged gain more from childcare in terms of their cognitivedevelopment than do other children (see, for example, Desai, Chase-Lansdale, and Michael, 1988; Vandell and Ramanan,1992; Caughy, DiPietro, and Strobino, 1994).
The fourth point is that one must be clear about what outcomes one cares about. To a large extent, the outcomes one tracks willdepend on the type of intervention being considered, the time at which it was delivered, and the type of children who receivedit, but it is important to remain open to unanticipated outcomes. Thus, in tracking the effects of early childcare, it is natural tofocus on issues of separation and attachment, but it would be useful to look at later social and cognitive outcomes as well.
1 See Carnegie Task Force on Meeting the Needs of Young Children (1994) and Shore (1997).2 To narrow the scope of this paper, I am specifically referring to childcare programs rather than to childcare policy more generally. There is a large literature
on the effects of childcare costs on women’s employment. For recent reviews, see Anderson and Levine (1998) and Han and Waldfogel (1998).3 Nor is there agreement on how to define quality of childcare. Childcare advocates tend to point to structural features of childcare programs such as the group
size, child-staff ratio, and health and safety requirements, while parents tend to look for a caregiver who is warm and sensitive, and conveniently located.Researchers try to measure both types of characteristics, as well as the continuity and stability of the care.
4 For recent evidence on this point, see Burchinal et al (1998) and the NICHD Early Child Care Network (1998).5 Very few studies have examined differences in outcomes associated with differences in timing within the first year of life. Baydar and Brooks-Gunn (1991)
is an important exception.
104
And, in assessing cognitively-orientated programs for older pre-schoolers, it makes sense to look at school outcomes but it isalso important not to lose sight of other outcomes that may be affected. Implicit in this discussion is the notion that it makessense to look at long-term as well as short-term outcomes, and at potential benefits for society as a whole in addition to thosethat may accrue to the child and his or her family.
With these ground rules in mind, let us now turn to the evidence on the potential benefits, and the potential ill effects, of earlychildhood interventions.
Potential benefits
We now know a good deal about what types of interventions at what time can have positive effects for what types of childrenand in what respects. Much of the evidence comes from the United States so the summary that I present will have a veryAmerican flavour.6 I will have more to say on this point later.
There have been several excellent reviews of the U.S. research on early childhood interventions and outcomes. The mostrecent, and the most useful for the purposes of this paper, is the RAND study which rigorously assessed nine early interventionprograms (Karoly et al, 1998).7 In order to be included in the RAND review, studies had to meet high scientific standards; inparticular, they had to have used random assignment or other techniques to control for pre-existing differences betweentreatment and controls and they had to follow the treatment and control groups over time so that they could assess long-term aswell as short-term outcomes. The results of the RAND review, summarized in Table 1, show that well-designed earlyintervention programs can make a positive difference in the lives of children. The results also show that the effects ofprograms vary by what specific type of program was offered. Eight of the nine programs were cognitively oriented and all ofthese programs were successful at raising children’s cognitive test scores or school achievement as measured by higher IQscores, higher school achievement test scores, less time in special education, better grades, less grade repetition, or higher ratesof graduation from high school. But the gains of these programs were not limited to cognitive outcomes. The High/ScopePerry Pre-School Project, for instance, led to higher employment, earnings, and income; it also led to lower rates of crime anddelinquency, as did two other programs (the Syracuse FDRP and the Chicago CPC programs). Interestingly, although mostprograms were child-focused, many were successful at changing parents’ behaviours in positive ways: the Elmira PEIP homevisiting program reduced abuse and neglect and also reduced parental welfare use; the Houston PCDC and the IHDP homevisiting and day care programs improved mother-child interaction and the HOME score (an index of how well the homeenvironment promotes child development); the Syracuse FDRP home visiting and day care program and the CarolinaAbecedarian program raised mothers’ level of education; the Carolina Abecedarian and IHDP programs raised maternalemployment; and the Chicago CPC day care and follow-through program raised parents’ involvement in their child’s school.8
Some of these effects on parents were intended but most were not.
Program outcomes varied by when services were delivered. In general, programs that intervened earlier and that were moreintensive (such as Carolina Abecedarian and IHDP) had stronger effects than those that intervened later and less intensively.Programs (such as Carolina Abecedarian and the Chicago Child-Parent Centers) that included a follow-through componentwere more successful at sustaining gains than those that didn’t. Consistent with prior research, some programs were morebeneficial for higher-risk children. For instance, the IHDP program produced the greatest IQ gains for the children with theleast educated parents.
The Rand study did not include Head Start because no Head Start evaluation met the Rand criteria for scientific rigour.However, Head Start is an important example: it is the single largest American childcare program and probably the best known.Early studies of Head Start concluded that the program had positive effects on children’s cognitive abilities and schoolachievement but these effects seemed to “fade out” over time ( McKey et al, 1985). However, the most recent evidence onHead Start reveals a more nuanced story (Lee et al, 1990; Currie and Thomas, 1995, 1996a, and 1996b). Children whoattended Head Start have higher test scores at the end of the program than siblings who stayed at home or attended anothertype of pre-school. Head Start children are also more likely to be immunized than siblings who stayed home. While the testscore effects for African-American children fade out fairly rapidly, perhaps because they go on to attend poor schools, theeffects for white and Hispanic children are longer-lasting. White and Hispanic children who attended Head Start have highertest scores at age 10 than comparable children who did not attend Head Start. White Head Start children are also less likely tohave repeated a grade by age 10 than comparable white children who did not attend Head Start.
Head Start continues to enjoy broad public and bipartisan support in the United States, and the program is now being expandedin two directions. First, Early Head Start is now delivering services to children under the age of three, reflecting the emphasison interventions in the first three years of life (and also reflecting the fact that older pre-school age children are increasinglylikely to be served by the public schools or other pre-schools). Second, Head Start Follow-Through programs are nowfollowing Head Start children into the school years, to see whether Head Start gains can be better maintained if follow-throughservices are provided.
6 For evidence on Sweden, see Andersson (1989) and Hwang (1990).7 See also recent reviews by Barnett (1995), Crane (1998), Ramey and Ramey (1998a, 1998b, and in press).8 Unfortunately, not all programs tracked parental outcomes, and those that did tended to track outcomes for mothers only. Thus, we do not know very much
about the effectiveness of early childhood interventions in changing the behavior of fathers.
105
Potential ill effects
There is also a fairly large body of research on the potential ill effects of early childhood interventions, although this researchhas tended to focus on a very narrow question, namely, whether maternal employment and early childcare – childcare begun inthe first year of life – have adverse outcomes for children. Much of the emphasis in this line of research has been onsocioemotional rather than cognitive outcomes, with a particularly vigorous debate about attachment. Several studies founddifferences in attachment between children who had been in early childcare and those who had not, but experts disagreed abouthow to interpret these results. If children who had been in early childcare engaged differently with their mothers, this might bea symptom of attachment problems (Belsky, 1988) or it might be a mature, adaptive response to the child care experience(Clarke-Stewart, 1988). Nor was it clear how such attachment differences might affect later outcomes.
This line of research, and the associated debate, dominated the childcare research agenda in the United States for many years.9
Only recently has it given way to interest in how specific types of child care early in a child’s life affects outcomes for specifictypes of children.
The progress in this area, at least in the U.S., is to a large extent a result of the formation of the NICHD early childcarenetwork. This unprecedented initiative brings together many of the country’s leading developmental psychologists, includingprominent representatives from both sides of the attachment debate, in a unique national longitudinal study of the effects ofearly childcare on child outcomes. Results from this study, which is still ongoing, are shown in Table 2. These results suggestthat one can not make sweeping conclusions about whether early childcare harms, or helps, children; rather, the effects of earlychildcare on a child’s attachment, child-mother interactions, and cognitive and behavioral outcomes depend critically on thecharacteristics of that care (including the quality of the care, its continuity, and the number of hours that the child is in care)and the characteristics of the child and family.10 Thus, increasingly, interest is shifting from the question of whether earlychildcare (or maternal employment) harms children to the question of what types of early childcare can be most helpful forwhat types of children.
What don’t we know about early childhood interventions and outcomes
In this concluding section, I want to particularly focus on what we don’t know about early childhood interventions andoutcomes in Britain. Much of the evidence I have cited comes from the United States which probably at least in part reflectsmy lack of knowledge about the British research base but also reflects the smaller size of that base.11 I want to focus on twoknowledge gaps in particular.
One, we don’t know enough about who is minding the children while mothers work in Britain. The labour force participationof women with young children, and especially those with infants, has increased sharply over the past few decades and is likelyto increase further in future.12 This trend presents both a challenge and an opportunity, and the outcomes for children willdepend to a large extent on the type and quality of the care they receive. Yet we know very little currently about what forms ofchildcare these mothers are using, and the quality of that care.13 Nor do we know which children begin care early, how youngthey are, and how many hours a week they are in care. Before we can begin to analyze the effects of childcare on outcomes forthese children, we need to understand who they are, when they are beginning care, and what types of care they are in.
Two, we don’t know enough about the effects of childcare and other early childhood interventions as delivered in Britain onoutcomes for children. Although we can learn a great deal from carefully conducted research in other countries, we do need tobe careful to compare like to like. We noted earlier that childcare is very heterogeneous, and of course there is even morevariation across countries than there is within them. Moreover, the effects of childcare may also be sensitive to the broaderpolicy context. For instance, we have seen in recent research that the long-run effects of pre-school intervention may dependon how supportive the child’s later school is and on whether follow-through programming is provided. Thus, research onBritish children, receiving British early childhood interventions and then entering British schools, is essential if we are to learnwhich early childhood interventions would be most effective and whether follow-through programming will be necessary toensure that effects do not fade out over time.
In summary, we now have enough evidence to conclude that early childhood interventions can make a difference, but we needto learn more about what types of childcare children in Britain are currently using and about what types of childcare and otherearly childhood interventions, delivered at what time and for which children, would achieve the best outcomes fordisadvantaged children in Britain.
9 There has been a very active debate on these topics in Britain as well (McGurk et al, 1993). Studies in Britain have produced mixed results about earlychildcare and socioemotional development; for instance, Osborn and Milbank (1987) found negative effects but Melhuish and Moss (1991) didn’t. Theresults for cognitive development have been more consistently positive; for instance both Osborn and Milbank (1987) and Melhuish and Moss (1991) reportpositive effects, as do recent reviews by Ball (1994) and Zoritch and Roberts (forthcoming), but see also Morgan (1996).
10 Early results from the NICHD study of early child care, which is following a total of 1364 children from 10 sites across the U.S., have been reported by theNICHD Early Child Care Research Network (1996, 1997, 1998 and in press).
11 For recent reviews of the British research, see Oliver, Smith, and Barker (1998) and Sylva (1994). See also the recent research by Francesconi and Ermisch(1998a, 1998b and 1998c) on the effects of maternal employment on later child outcomes.
12 The share of infants with working mothers has risen from 20% in 1981 to 36% in 1990 to 47% in 1997 (Gregg and Wadsworth, 1998). Current policyinitiatives such as increased rights to parental leave, the childcare tax credit, and the New Deal for lone parents are expected to lead to further increases inthe share of women working and using childcare while their children are very young.
13 In the 1991-92 GHS, 46% of families with a child under the age of one used some form of non-parental child care, with about 16% using unpaid informalcare, 10% using nurseries, and 20% using other forms of paid care. The share of infants in care has probably risen a good deal since then.
ReferencesAnderson, P. and P. Levine (1998). “Child Care and Mother’s Employment Decisions”. Paper presented at the Conference on Labour Markets and Less-Skilled Workers, Washington, D.C., November 5, 1998.Andersson, B.-E. (1989). “Effects of Public Day Care: A Longitudinal Study”. Child Development, 60, 857-867.Ball, C. Start Right: The Importance of Early Learning. London: Royal Society of the Arts. Barnett, W.S. (1995). “Long-Term Effects of Early Childhood Programs on Cognitive and School Outcomes”. The Future of Children, 5(3), 25-50.Bates, J.E., D. Marvinney, T. Kelly, K.A. Dodge, D.S. Bennett, and G.S. Pettit (1994). “Child-Care History and Kindergarten Adjustment.” DevelopmentalPsychology, 30, 690-700.Baydar, N. and Brooks-Gunn, J. (1991). “Effects of Maternal Employment and Child Care Arrangements in Infancy on Preschoolers’ Cognitive and BehavioralOutcomes: Evidence from the Children of the NLSY.”Developmental Psychology, 27, 918-931.Belsky, J. (1988). “The ‘Effects’ of Infant Day Care Reconsidered.” Early Childhood Research Quarterly, 3, 235-272.Belsky, J. and D. Eggebeen (1991). “Early and Extensive Maternal Employment/Child Care and 4-6 Year Olds Socioemotional Development: Children of theNational Longitudinal Survey of Youth.” Journal of Marriage and the Family, 53, 1083-1099.Blau, F. and A. Grossberg (1990). “Maternal Labour Supply and Children’s Cognitive Development.” NBER Working Paper.Brooks-Gunn, J., F. Liaw, and P. Klebanov. (1992). “Effects of Early Intervention on Cognitive Function of Low Birth Weight Preterm Infants.” The Journalof Pediatrics, 120(3), 35-358.Brooks-Gunn, J., C. McCarton, P. Casey, M. McCormick, C. Bauer, J. Bernbaum, J. Tyson, M. Swanson, F. Bennett, D. Scott, J. Tonascia, and C. Meinert.(1994). “Early Intervention in Low-Birth-Weight Premature Infants: Results Through Age 5 Years From the Infant Health and Development Program.”Journal of the American Medical Association, 272(16), 1257-1262.Brooks-Gunn, J., P.K. Klebanov, F. Liaw, and D. Spiker (1993). “Enhancing the Development of Low-Birthweight, Premature Infants: Changes in Cognitionand Behavior over the First Three Years.” Child Development, 64, 736-753.Burchinal, M. R., J.E. Roberts, R. Riggins, Jr., S.A. Zeisel, E. Neebe, and D.Bryant (1998). “Relating Quality of Center Child Care to Early Cognitive andLanguage Development Longitudinally”, paper presented at the International Conference on Infant Studies, Altanta, Georgia.Carnegie Task Force on Meeting the Needs of Young Children (1994). Starting Points: Meeting the Needs of Our Youngest Children. New York: CarnegieCorporation.Caughy, M.O., J. DiPietro, and D. Strobino (1994). “Day-Care Participation as a Protective Factor in the Cognitive Development of Low-Income Children.”Child Development, 65, 457-471.Clarke-Stewart, K.A. (1988). “‘The ‘Effects’ of Infant Day Care Reconsidered’ Reconsidered: Risks for Parents, Children, and Researchers.” Early ChildhoodResearch Quarterly, 3, 293-318.Crane, J. (editor) (1998). Social Programs That Work. (New York: Russell Sage Foundation).Currie, J. and D. Thomas (1995). “Does Head Start Make a Difference?” American Economic Review, 85(3), 341-364.Currie, J. and D. Thomas (1996a). “Head Start and Cognition Among Latino Children”. Mimeo, University of California at Los Angeles, Los Angeles, CA.Currie, J. and D.Thomas (1996b). “Could Subsequent School Quality Affect the Long Term Gains from Head Start?” Working Paper, National Bureau ofEconomic Research, Cambridge, MA.Desai, S., L. Chase-Lansdale, and R. Michael (1988). “Mother or Market?: Effects of Maternal Employment on Cognitive Development of Four Year OldChildren.” Demography, 26, 545-561.Field, T. (1991). “Quality Infant Day-Care and Grade School Behavior and Performance.” Child Development, 62, 863-870.Francesconi, M. and J. Ermisch (1998a). “Mother’s Behaviour and Children’s Achievement”. Working Paper Number 98-3, ESRC Research Centre on Micro-social Change, University of Essex.Francesconi, M. and J. Ermisch (1998b). “Mother’s Employment, Lone Motherhood and Children’s Achievements as Young Adults”. Working Paper Number98-9, ESRC Research Centre on Micro-social Change, University of Essex. Francesconi, M. and J. Ermisch (1998c). “Family Structure and Children’s Achievements”. Mimeo, ESRC Research Centre on Micro-social Change,University of Essex. Gregg, P. and J. Wadsworth. (1998). “Gender and the Labour Market”. Forthcoming in P. Gregg and J. Wadsworth, The State of Working Britain.Han, W.J. and J. Waldfogel (1998). “Child Care Costs, Regulations, and the Employment of Married and Unmarried Mothers”. Paper presented at the AnnualMeeting of the Population Association of America, Chicago, IL, April 1998.Howes, C. (1988). “Can the Age of Entry into Child Care and the Quality of Child Care Predict Adjustment in Kindergarten?” Developmental Psychology,1990, 26(2), 292-303.Howes, C. (1990). “Relations Between Early Child Care and Schooling.” Developmental Psychology, 24(1), 53-57.Hwang, P. (1990). “Day Care in Sweden”. In P. Moss and E. Melhuish (eds) Current Issues in Day Care for Young Children. London: HMSO.Karoly, L.A., P.W. Greenwood, S.S. Everingham, J. Hoube, M.R.Kilburn, C. P. Rydell, M. Sanders, and J. Chiesa (1998). Investing in Our Children: What WeKnow and Don’t Know about the Costs and Benefits of Early Childhood Interventions. (Santa Monica, CA: RAND).Lee, V., J. Brooks-Gunn, E. Schnur, and F-R. Liaw (1990). “Are Head Start Effects Sustained? A Longitudinal Follow-Up comparison of DisadvantagedChildren Attending Head Start, No Preschool, and Other Preschool Programs.” Child Development, 61, 495-507.McGurk, H. M. Kaplan, E. Hennessy, and P. Moss (1993). “Controversy, Theory, and Social Context in Contemporary Day Care Research”. Journal of ChildPsychology and Psychiatry, 34, 3-23.McKey, R.H., L. Condelli, H. Ganson, B.J. Barrett, C. McConkey, and M.C. Plantz (1985). The Impact of Head Start on Children, Families, andCommunities. Washington, D.C.: U.S. Government Printing Office.Melhuish, E.C. and P. Moss (1991). Day Care for Young Children: International Perspectives. London: Tavistock/Routledge.Mooney, A. and A. Munton (1997). Research and Policy in Early Childhood Services: Time for a New Agenda. London: Thomas Coram Research Unit.Morgan, P. (1996). Who Needs Parents? London: Institute for Economic Affairs.NICHD Early Child Care Research Network (1996).“Characteristics of Infant Child Care: Factors Contributing to Positive Caregiving”. Early ChildhoodResearch Quarterly, 11, 269-306. NICHD Early Child Care Research Network (1997). “The Effects of Infant Child Care on Infant-Mother Attachment Security: Results of the NICHD Study ofEarly Child Care”. Child Development, 68, 860-879. NICHD Early Child Care Research Network (1998). “When Child-Care Classrooms Meet Recommended Guidelines for Quality”. Paper presented at ameeting on Child Care in the New Policy Context, U.S. Department of Health and Human Services, Bethesda, Maryland, April 30, 1998.NICHD Early Child Care Research Network (in press). “Early Child Care and Self-Control, Compliance, and Problem Behavior at 24 and 36 Months”.Forthcoming in Child Development.Oliver, C., M. Smith, and S. Barker (1998). “Effectiveness of Early Interventions”, supporting paper for the Comprehensive Spending Review: Cross-Departmental Review of Provision for Young Children, HM Treasury, July 1998.Osborn, A.F. and J.E. Milbank (1987). The Effects of Early Education: A Report from the Child Health and Education Study. Oxford: Clarendon Press.Ramey, C.T. and S.L. Ramey (1998a). “Early Intervention and Early Experience”. American Pyschologist, 53, 109-120.Ramey, S.L. and C.T. Ramey (1998b). “The Effects of Early Childhood Experiences on Developmental Competence”. Paper presented at the Conference onInvesting in Children, Columbia University, New York, New York, October 15-17, 1998.
106
Ramey, C.T. and S.L. Ramey (in press). Right from Birth: Building Your Child’s Foundation for Life. New York: Goddard Press, forthcoming.Shore, R. (1997). Rethinking the Brain: New Insights into Early Development. New York: Families and Work Institute.Smith, J. (1994). “Maternal Employment and the Young Child.” Ph.D. thesis, Columbia University.Sylva, K. (1994). “The Impact of Early Learning on Children’s Later Development” in C. Ball Start Right: The Importance of Early Learning. London:Royal Society of Arts.Vandell, D., V.K. Henderson, and K.S. Wilson (1988). “A Longitudinal Study of Children with Day-Care Experiences of Varying Quality.” ChildDevelopment, 59, 1286-1292.Vandell, D. and J. Ramanan (1992). “Effects of Early and Recent Maternal Employment on Children from Low-Income Families.” Child Development, 63,938-949.Zoritch, B. and I. Roberts (forthcoming). The Health and Welfare Effects of Preschool Daycare: A Systematic Review of Randomised Controlled Trials.London: Institute of Child Health.
107
108
Table 1
The Effects of Early Childhood Interventions: Selected U.S. Studies
109
Table 2
Results from the NICHD Study of Early Child Care
* Differences statistically significant for the high-risk group only.** Differences statistically significant for girls only.*** Differences statistically significant for heavier LBW children only.
Source: Karoly et al, 1998, Tables 2.1, 2.2, and 2.3.
• Childcare per se neither helps nor harms attachment.
For children whose mothers are sensitive caregivers, childcare has no effect on attachment. For children whosemothers are not sensitive, care matters: high quality care leads to more secure attachment, while poor quality care,more than 10 hours per week of care, or more than 1 care arrangement by age 15 months leads to less secureattachment.
• Quality of care has an effect on mother-child relationships.
Higher quality care predicts greater involvement and sensitivity by the mother at 15 and 36 months and more positiveinteractions at 36 months. Low-income mothers using high-quality child care have more positive interactions withtheir children at age 6 months than those who do not use care or who use lower-quality care.
• The quantity of care seems to matter as well.
Longer hours of care in the first six months are associated with lower maternal sensitivity and less positive interactionsat 36 months. Longer hours of care are also associated with more reported behavior problems at age 2. But, child andfamily characteristics are more important.
• The quality of child care in the first three years of life affects children’s cognitive and language development.
The higher the quality of care – in terms of language stimulation and the type of interactions between the child andcaregiver – the higher the child’s language skills at 15, 24, and 36 months. Higher quality care also is associated withcognitive development at age 2 and school readiness at age 3. Children in day care centers that meet quality standardsacross all four domains assessed – child-staff ratio, group size, teacher training, and teacher education – have betterlanguage comprehension and school readiness, and fewer behavior problems, than children whose day care centers failto meet the standards in all four domains.
110
‘Who divorces? and the legacy of parental divorce’
Kathleen Kiernan, Department of Social Policy and Centre for Analysis of Social Exclusion, London School of Economics
Who divorces?
The subject matter of this presentation was Divorce and Family Breakdown and we started by posing the question WhoDivorces? as this has important ramifications for the study of the impact of divorce on children’s lives. Divorce is used as aninclusive short hand term which includes the break-up of marital and cohabiting unions.
In a recent research report (Kiernan and Mueller, 1998) we posed two broad questions: what are the characteristics of thecurrently divorced; and who divorces? In the first part of this study we used data from the Family Resources Survey toidentify the characteristics of the divorced population. In the second part we used two longitudinal studies, the BritishHousehold Panel Survey (BHPS) and the National Child Development Study (NCDS) to address the question “Who divorces”.
The BHPS allowed us to examine this issue for individuals and couples of all ages whereas the data from the NCDS allowed usto examine background factors from childhood and adolescence associated with partnership dissolution in adulthood. Anumber of insights emerged from our longitudinal analyses as well as from the cross-sectional analysis of the FamilyResources Survey.
We found that unemployment, reliance on state benefits and disability featured as characteristics of the currently divorced inthe FRS and these factors, along with financial difficulties, were also found to be important precursors of divorce from ouranalysis of the BHPS. This suggests that poor economic and physical well-being may be important stressors in a relationshipand that the selection of vulnerable groups into divorce may be an important aspect of the poverty observed amongst thepreviously partnered, as well as the deprivation that may be a by-product of the divorce itself.
There was also evidence from both the BHPS and the NCDS of an association between emotional factors and subsequentpartnership breakdown. The analysis of the BHPS showed that men and women with lower psychological well-being weremore likely to divorce in the ensuing few years and analysis of the NCDS data suggested that pre-existing emotional problemsin childhood were important signposts for subsequent partnership breakdown. Again these two findings spoke to the possibilityof selection effects as well as emerging emotional problems post-partnership being implicated in the lower emotional well-being of the divorced.
The legacy of parental divorce
Our research and that of others has shown divorce to be more likely to occur amongst couples with personal, social andeconomic problems. Thus one of the challenges in assessing the legacy of divorce for children is being able to sort out theconditions that lead couples to separate and the potential effects on children from the consequences of the dissolution itself.
For example, the non-random nature of the divorcing population implies that the effects of factors that existed prior to thedivorce, for example, poverty, could be confused with its consequences. Additionally the selective nature of the population ofchildren who experience parental divorce may lead to an over-stated impression of the effects of divorce by conflating pre-existing differences amongst children from disrupted families as compared with those from non-disrupted ones with the falloutfrom marital dissolution.
To begin to address this challenge we pursued three questions in our study of the legacy of parental divorce on social andeconomic and family experiences in adulthood (Kiernan, 1997). For this study we used the NCDS data and posed the followingquestions.
• To what extent does parental divorce during childhood (0-16 years) have long-term consequences for educationalattainment, economic circumstances, partnership formation and parenthood behaviour in adulthood?
Specifically we examined the following outcomes:
Educational outcomes Qualifications attained by ages 23 and 33;
Economic outcomes: Earnings and Household Income at age 33. Receipt of benefits at age 33. Unemployment at age 33and experience since leaving school;
Housing outcomes: In social housing and of experience in homelessness;
Demographic outcomes: Timing of first and type of first partnership; timing of parenthood and partnership context andpartnership dissolution;
111
The control variables included in our analyses were measures of financial difficulties, test scores (reading and maths) andbehavioural problem scores at ages 7 and 16 and social group at age 7.
• Secondly, when child and family characteristics before divorce are taken into account, does the link between the divorceand adult experiences become weaker?
To address this question we divided the sample who had experienced parental divorce during childhood into two groups: thosewhose parents divorced when they were under age 7 years and those who experienced parental divorce when they werebetween 9 and 16 years of age. We chose this strategy for a number of reasons. Firstly, age 9 unambiguously post-dates theage 7 interview which allows us to assess whether attributes that chronologically precede parental separation are implicated inlater life experiences. Only children who were living with both their parents at age 9 were included in the analysis so that anychanges in educational performance or behaviour that might be associated with parental separation should be excluded.However, we recognise that parental conflict prior to divorce may also have, for example, a dampening effect on a child’s testscores at age 7 or increase behavioural problems, but we have no measure that allows us to control for this situation. We alsoexamined the 7 year old attributes for those children who experienced parental divorce before age 7 to assess whether theymoderated or amplified subsequent behaviour – but for this group we do not know whether differences were due to the after-math of the separation or pre-disruption factors.
• Thirdly, if parents stay together until their children are grown up before separating does this reduce the impact of divorceon their adult children’s lives?
To address this question we compared the experiences of children who experienced parental divorce at ages 20 and older withthose who experienced parental divorce during childhood (0-16 years).
Here we highlight some of the findings from the analysis.
Educational attainment
First we focus in on the group with no qualifications. By the time they were 33, those who had experienced parental divorce aschildren (16 and under) were almost twice as likely as others to lack formal qualifications: 20 per cent compared with 11 percent, and children who experienced parental divorce were under-represented amongst graduates. We also compared age 23 and33 qualifications, and found similar proportions of children from divorced and non-divorced backgrounds had improved ontheir age 23 qualifications. However, it was the young people who already had qualifications that were most likely to improvetheir position. Children who experienced parental divorce during childhood were less likely than their peers to have anyqualifications at age 23 and this continued to be the position at age 33. This suggests that remedial action with respect toeducational underachievement amongst vulnerable groups may have greater returns if executed during the school years than inlater years and that particular attention needs to be given to recruiting unqualified adults for further training and education.
Turning to the question of whether the lower educational attainment of children arises from divorce or to prior factors, Table 1shows the odds ratios of having no qualifications relative to children brought up with both parents. The analysis was carriedout in a series of steps to clarify whether particular factors were more or less important in lowering the chances of having noqualifications. So we entered the individual factors, financial hardship, test scores, behaviour scores and social class at age 7separately and then in combinations. It is fairly clear from this table that financial hardship in the family at age 7 substantially
112
Table 1: Odds ratios for effects of parental separation by age of child on having no qualifications in adulthood
Men Women
9-16 years 0-6 years 0-16 years 9-16 years 0-6 years 0-16 years
Baseline 1.44* 2.11*** 1.62*** 2.04*** 2.37*** 2.16***
Financial hardship age 7 0.91 1.80* 1.19 1.30 1.53 + 1.44**
Behaviour scores age 7 1.20 1.38 1.20 1.83*** 1.52 1.65***
Test scores age 7 1.34 1.71** 1.40** 1.89*** 2.11*** 1.88***
Social group at 7 1.33 2.09*** 1.53** 2.05*** 2.41*** 2.14***
All age 7 factors 0.88 1.63 1.10 1.27 1.16 1.24
All age 7 and age 16 factors 0.63 1.19 0.80 1.12 1.04 1.11
Source NCDS.
Notes: + significant at the .10 level; * significant at the .05 level; ** significant at the .01 level; *** significant at the .001 level.
attenuates the difference between children from divorced families and those from intact families with respect to having noqualifications in adulthood. For example, the baseline odds ratio of 2.04 is reduced to 1.3 in the case of the women and in thecase of the men where the relationship is somewhat weaker from 1.44 to 0.91. Controlling for the other individual factorsnamely: the child’s cognitive test scores at 7 and behavioural problems led to some reduction in the odds of having noqualifications, but in the case of women controlling for these factors did not significantly attenuate the observed differencebetween women from intact and disrupted families. Turning to consider the children who experienced parental divorce priorto age 7 we see amongst the women that the chances of having no qualifications as compared with those who had notexperienced a parental divorce by age 7 is also much less when we control for financial well-being and behaviour problemscores at age 7. Amongst the men the differences between those from intact and disrupted families is much less when we takeinto account level of behavioural problems at age 7. Where parental divorce occurred prior to age 7 we cannot disentanglewhether differences in educational outcomes are due to parental separation or to selection or a combination of selection and anamplification of financial problems, but the findings with respect to the post age 9 group suggest that selection may be animportant element, but not an exclusive one, in the interplay between the divorce process and the educational attainment ofchildren. Poverty and behaviour problems reduce educational success and parental divorce can amplify both.
Adult income and employment
Turning to adult income and experience of unemployment we found that adult men who had experienced divorce duringchildhood had broadly similar earnings and family incomes to contemporaries whose families had remained intact.
However, men with a background of parental divorce were at greater risk of unemployment at age 33 and of having beenunemployed on more than one occasion since they left school. The odds of unemployment as an adult were greatest for boyswhose parents had separated before they were 7 years old and introduction of financial hardship at age 7 reduced the oddsslightly but they remained high and significant and this applied to all the individual age 7 controls and in combination.However, amongst the group of boys who experienced divorce between age 9 and 16 controlling for financial hardship at age 7resulted in a reduction in the odds of unemployment to a level that was no longer statistically significant among men whoseparents divorced during their later childhood. This suggests, as with our findings on qualifications, that family circumstancesprior to divorce may be implicated in children having adverse employment experiences in adulthood. The income andemployment histories of women are complicated by the timing of motherhood. Women who had experienced divorce aschildren were more likely to have begun child-bearing at an early age and this, in turn, affected their ability to earn. In general,women whose parents had divorced had lower earnings than others and were more likely to be receiving Income Support thanwomen from intact families – although the latter link was relatively weak.
Housing
Adults who experienced parental divorce in childhood were more likely than others to be living in property rented from a localauthority or housing association and less likely to be home owners. Women from divorced families were especially likely to beliving in social housing at age 33. Once again, the links between a divorce background and living in social housing as an adultappeared largely to be indirect. Taking financial hardship, behavioural problems and other factors present at age 7 into accountreduced the odds for both men and women.
First partnerships and parenthood
With respect to the demographic outcomes we found that men and women who experienced parental separation duringchildhood were more likely to form cohabiting rather than marital unions, and to cohabit or marry at a younger age than thosewhose parents stayed together. Men and women from divorced families were also more likely to become parents at a youngage. Partnership break up was also greater amongst children who had experienced parental divorce. But across all thesedemographic outcomes, controlling for the influence of childhood and adolescent background factors produced no or only amodest reduction in the odds among men and women whose parents had divorced.
Post-childhood divorce
Surprisingly, little is known about the impact of parental divorce that occurs after childhood, although this is an increasinglycommon experience. In our study we compared outcomes for 33-year old adults who were aged over 20 when their parentsdivorced with those who parents separated during childhood and with those whose families remained intact. Women from laterdivorcing families were similar to their peers from intact families in terms of their educational attainment and economicsituation in adulthood but there were indications that men these families still tended to experience some social and economicdisadvantage. Both men and women from later divorcing families were similar to those brought up with both parents in theirodds of forming partnerships or becoming parents at a young age. However, both men and women whose parents divorcedwhen they were adults were more likely than those whose parents had not divorced to cohabit and to experience the break-upof their own partnerships or marriages.
Conclusions
The message from this research is that there are no simple conclusions to be drawn about the way that parental divorce duringchildhood is linked to behaviour in adulthood. The connection – when it exists – depends on which aspects of adult life areexamined. It also depends on whether the children in question are boys or girls and, to some extent, on their age at the time of
113
separation. Pre-divorce factors, especially financial hardship, played an important part in explaining part of the increased oddsthat children whose parents divorced would lack qualifications, be unemployed or be living in social housing as adults. Pre-divorce circumstances were less influential in accounting for why children who experienced parental divorce differed in theirpersonal relationships and parenthood behaviour in adulthood. It is undoubtedly the case that children benefit from beingraised in an emotionally and economically secure two-parent family. But if that is not possible, there is evidence from thisstudy which suggests that if we are concerned with children’s long-term welfare then we should perhaps be as concerned withthe conditions that precede, and may lead to parental separation such as poverty, economic uncertainty and other stressors, aswith its consequences.
ReferencesK. Kiernan and G. Mueller 1998. Who are the Divorced and Who Divorces? CASEpaper No7 ESRC Centre for the Analysis of Social Exclusion, LondonSchool of Economics K. Kiernan 1997. The Legacy of Parental Divorce: Social, economic and demographic experiences in adulthood. CASEpaper No1 ESRC Centre for theAnalysis of Social Exclusion, London School of Economics.
114
Intergenerational and Life-Course Transmission of Social Exclusion:Influences of Childhood Poverty, Family Disruption, and Contact with the Police
John Hobcraft, Centre for Analysis of Social Exclusion (CASE) and Department of Social PolicyLondon School of Economics & Political Science
Questions Posed
How far is social exclusion and disadvantage transmitted from parents to their children and from childhood into adulthood?
In particular, how far do childhood experiences of poverty, family disruption, and contact with the police link to adultoutcomes?
What associations are there for a range of other parental and childhood factors – social class of origin, social class duringchildhood, housing tenure, father’s and mother’s interest in schooling, ‘aggression’, ‘anxiety’, and ‘restlessness’, andeducational test scores?
How do these factors link to outcomes by age 33, including three indicators of demographic behaviour, one of psychologicalwell-being, three of welfare position, two of educational qualifications and three of economic position?
Which childhood factors have a general influence on adult exclusion and are there specific ‘inheritance’ patterns?
This study uses data from the National Child Development Study (NCDS), a longitudinal study of children born in 1958, toexamine these questions. Innovative methods are used to maximise the extent to which available information is used, includingretention of missing information during childhood. Emphasis is placed upon the cumulation of childhood experiences, usuallyat ages 7, 11, & 16. Let’s now turn to the wide range of childhood antecedents and adult outcomes considered.
Adult Outcomes by age 33 – restrict attention to ten negative outcomes for today
Early parenthood – Father by 22; Mother by 20Extra-marital birthsThree or more partnerships
Malaise (at risk of depression)
Social housingReceipt of benefitsHomelessness
No qualificationsLow income – own for men; household for womenUnemployment (men only)
‘Focal’ childhood variables
PovertyAt ages 7, 11 & 16Reports of serious financial difficultiesReports of free school meals (11 & 16)
PoliceBy age 16‘in trouble with the police’ (teacher)‘contact with police/ probation’ (teacher)‘contact with police/ probation’ (parent)
Family Typecovers experience at ages 0, 7, 11 &16both natural throughoutboth natural, some missing data
born outside marriageever in care or fostering
115
divorce, no remarriageother one-parent, no remarriagedivorced & remarriedother one-parent & remarried
These childhood focal variables are themselves quite closely interrelated• 44% of the poorest boys had contact with the police by age 16, compared with only 13% for the non-poor.• 47% of children with divorced lone-parents experienced childhood poverty, as defined by family financial difficulties and
free school meals, compared with only 8% in intact two-parent families.
‘Control’ childhood variables
Common approach – all summarised into four groups (low, middle, high & no information) to maximise information used(including partial or completely missing)
Social class of origin – two grandfathers father at birth Social class of father – ages 7, 11 & 16 Housing tenure – ages 7, 11 & 16 Father’s interest in schooling – 7, 11 & 16 Mother’s interest in schooling – 7, 11 & 16 ‘Aggression’ scores – 7, 11 & 16 ‘Anxiety’ scores – 7, 11 & 16 ‘Restlessness’ scores – 7, 11 & 16 Reading and maths tests – 7,11 & 16
Summary groups used:2/3 ‘adverse’1 adverse 0 adverse, 0 or 1 ‘beneficial’0 adverse, 2/3 beneficial All information missing
CASEbrief 8 (distributed at the meeting) contains quite a number of specific illustrations drawn from the larger study, many ofwhich I shall not repeat in this brief talk. I shall try to highlight some of the key issues that arise from the results, especiallythose that benefit from the broad compass of the childhood antecedents and the adult outcomes. The particular benefit of thisinterdisciplinary approach is that the associations found are more robust than usual, because of the depth and range of thecontrol variables, the cumulation of childhood experiences, the handling of missing data, and not least the causal priority of thechildhood measures.
The Pervasive Influence of the Focal Variables: Childhood Poverty, Family Disruption, and Contact with the Police
Childhood experiences of poverty, family disruption, and contact with the police are all powerful antecedents of adultdisadvantage. The Table shows how many of the ten negative male and nine negative female adult outcomes are stronglyrelated to the 12 childhood antecedents. The three focal variables and educational test scores prove to have the most pervasivestrong influences on adult outcomes by quite a margin:
• educational test scores to 15 of 19
• childhood poverty is clearly linked to 15 of 19, contact with the police to 15 of 19,
• family type to 14 of 19.
• the next most pervasive is father’s interest in schooling (8 of 19).
116
Numbers of clear associations of childhood variables with negative adult outcomes
Men Women Both sexes(of 10) (of 9) (of 19)
Tests 7 8 15Poverty 6 9 15Police 7 8 15Family Type 7 7 14
Dad Interest in School 3 5 8Mum Interest in School 2 4 6Aggression 2 3 5Social Class of Origin 1 4 5Social Class of Father 3 1 4Tenure 2 2 4Anxiety 3 1 4Restlessness 1 0 1
LARGE AND SPECIFIC INFLUENCES ON ADULT OUTCOMES
The next Table addresses a different question. It shows, for each childhood antecedent, the adult outcome with largest oddsratio. Specific continuities (in the sense that the childhood antecedent is directly transmitted into closely related adultoutcomes) among these largest odds ratios are picked out in bold type. I note that some specific continuities are not possible,since several of the childhood antecedents have no direct counterpart among the adult outcomes considered here (contact withpolice, social class of origin and social class of father during childhood, aggressiveness, and restlessness). What is remarkableis how frequently the largest odds ratio occurs for these specific transmissions, despite the wide range of adult outcomesconsidered.
Adult negative outcomes showing largest odds ratio for each childhood variable
Men Women
Childhood factor Outcome Odds Outcome OddsRatio Ratio
Poverty No qualifications 2.8 No qualifications 2.6
Police No qualifications 3.7 No qualifications 2.4
Family Type
Born-out-of-Wedlock Extra-Marital Birth 2.0 Extra-Marital Birth 2.5Ever in Care Malaise 2.1 Extra-Marital Birth 3.7
Divorce 3+ Partners 3.2 3+ Partners 2.3Remarriage Homelessness 2.8 Extra-Marital Birth 1.7
Tests No qualifications 45.9 No qualifications 26.8
Social Housing 4.3 Teenage mother 3.7Low Income 3.9 Social Housing 2.7Young Father 3.4
Dad Interest No qualifications 4.3 No qualifications 3.7Mum Interest Extra-Marital Birth 2.1 No qualifications 2.5
Tenure Social Housing 2.5 Social Housing 1.8
Social Class of Origin Malaise 1.5 Social Housing 2.4Social Class of Father Young Dad 2.5 Social Housing 1.4
Aggression 3+ Partners 1.8 Teenage mother 1.9Anxiety Malaise 1.6 Malaise 1.7Restlessness No qualifications 1.9 Homeless 1.7
117
SPECIFIC CONTINUITIES ONLY
Adult negative outcomes showing odds ratio for each childhood variable(including non-largest for poverty)
Men Women
Childhood factor Outcome Odds Outcome OddsRatio Ratio
Tests No qualifications 45.9 No qualifications 26.8
Dad Interest No qualifications 4.3 No qualifications 3.7Mum Interest
No qualifications 2.5
Born Out-of-Wedlock Extra-Marital Birth 2.0 Extra-Marital Birth 2.5
Divorce 3+ Partners 3.2 3+ Partners 2.3
Tenure Social Housing 2.5 Social Housing 1.8
Anxiety Malaise 1.6 Malaise 1.7
Poverty Any Benefits 1.6 Any Benefits 1.4
Low income 1.5 Low income 1.5Unemployment 1.4
Adult Outcomes
I now want to turn to a brief summary which looks at each of the Adult Outcomes in turn and highlights the main associationswith the various childhood antecedents, which draws heavily on CASEbrief 8. (Some more detail is shown in CASEpaper 15,Table 21, pages 59-61, which includes all childhood antecedents which are statistically significant for each adult outcome inrough order of the ‘strength of association’ for each output.) The highlights mentioned here cover only associations that aresubstantively large.
Young parenthood: poorly socialised girls appear more likely to become young mothers (contact with the police, incare/fostering or born out-of-wedlock, aggressive, low performance on educational tests, and lacking maternal interest inschooling are the factors with the highest odds). Young fathers are quite likely to have had contact with the police, to comefrom a lower social class, and to have performed poorly at school.
Extra-marital birthsare more frequent for children of either sex who were themselves born out of wedlock, for girls who werein care or fostered and, to a lesser extent and less consistently, for men and women who experienced post-birth familydisruption; there is also a heightened risk for those of either sex whose parents were less interested in their schooling and whohad been in contact with the police.
Multiple partnershipsare particularly common for men and women who had experienced parental divorce during theirchildhood, for ‘delinquents’ who had childhood contact with the police and for aggressive men.
Adult malaiseis more common for anxious children, poor educational performance as a child, delinquent girls, children whowere in care, poor children and aggressive and restless males.
Living in social housingas an adult is strongly associated with being in local authority housing as a child, coming from a lowersocial class background, poor performance on educational tests, being poor and delinquent.
Receipt of benefitsin adulthood is linked to poor educational testing as a child, poverty and delinquency.
Adult homelessnessis most common among those with step-parents in childhood, but also for delinquents, girls who were pooror in care, and boys who were not sons of owner-occupiers.
Adult unemploymentis more likely for boys who were delinquent, poor, in care, or had mothers who were not interested intheir schooling.
Educational failure is dramatically increased by lack of parental interest in schooling, by childhood poverty, and bydelinquency; girls who were in care or from lower social class of origin, and boys who were fidgety or restless also faileducationally.
118
Low income in adulthoodis related to poor performance at school and lack of paternal interest in schooling, more sharply formen; and, to a lesser extent, to childhood poverty for both sexes.
CONCLUDING REMARKS
Family disruption is most related to demographic outcomes and homelessness and social housing.
Childhood experience of care or fostering has pervasive association with all negative adult outcomes for females.
Educational test scores are powerful predictors of a wide range of adult outcomes.
Father’s interest in schooling is important and relatively more so than mother’s interest, although there is also a greaterreinforcing effect of mother’s interest for females.
There is a tendency for social and parental factors to be more related to adult exclusion for females and for external andstructural factors to be more related to exclusion for males.
There is a pervasive association of most negative adult outcomes with educational tests, poverty, contact with police andfamily type.
A large number of specific life-course and inter-generational continuities in transmission of social exclusion are identified,including education, social housing, out-of-wedlock births, partnership breakdown, anxiety, and poverty.
There is little doubt that social exclusion, as captured by the adult outcomes and childhood factors used here, is transmittedacross the generations and through the life-course. There is also little doubt that there are a large number of very specificcontinuities from childhood experience into adulthood. But it is essential to emphasise that all of the associations captured hereare just aggregate tendencies and in no sense determinist. Specific and general disadvantage during childhood is echoed inadulthood in specific and general forms. There is huge scope for many, if not most, individuals to escape from the patterns andtendencies observed. An important potential area for further research is to examine more closely the characteristics ofindividuals who escape the general tendencies. Such work will involve much more detailed examination of individual recordsand of clusters and combinations of childhood factors (and intervening experience during adulthood) than has so far beenpossible.
Another open question arises with respect to whether the timing of particular forms of disadvantage during childhood is crucialor at least further aids the understanding of how people reach adult outcomes. This study of pathways is again more complexand raises greater difficulties with respect to missing information.
Despite the general, but not complete, causal priority of our explanatory variables both in time and in their measurement, weremain cautious about attaching causality to the associations observed, regardless of the plausibility of the links. Without amuch more thorough understanding of pathways and protective factors it is extremely unwise to jump to facile policyimplications from this work. The problems studied here have been the focus of much social policy and governmentalintervention over the years. Just to give one interpretational difficulty, let us look at education.
Everyone agrees that improving the effectiveness of schooling is an important and desirable goal. But there are a number ofpossible policy levers which may have differing appeals to differing political persuasions: increasing parental interest inschooling; a sub-theme for some would then be to try to prevent family disruption and retain interested fathers, and for otherswould be to accept family breakdown but intervene to keep both parents more engaged. Making education super-efficient mayalso fail to prevent some of the associations examined here. Since all of our educational test scores are relative, dividing thescores into quartiles, it is clear that this distribution cannot change very much. Indeed, it is quite likely that the differentialswould sharpen rather than reduce with greater educational efficiency, since sorting on potential would become more efficient.
What is overwhelmingly clear from this work is the extent and pervasiveness of both specific and general continuities acrossthe generations and across the life-course in the transmission of aspects of social exclusion. Interpretation depends heavily onassumptions about causality, on debates about nature versus nurture, and on debates about structural constraints and individualopportunity. It is hardly surprising that we have not answered such thorny questions.
Next Steps using NCDS
Explore pathways and links of changes during childhood to process of social exclusion over the life courseMore on protective factorsOutcomes at 23 and the timing and sequencing of experiences in early adulthoodIntermediate outcomes – contact with police, educational performance, behavioural change etcLinks to participation in civil society
119
Session 5: Childhood poverty and family structure: Discussion
NCDS evidence on early years
Professor Bynner started off by reminding us about the sources used and looked at some broad and not so broad definitions andhow you needed to be certain about when interpretating this data what the definitions of poverty were all about and what theymeant.
He then went on to talk about cohort shifts, namely whether each cohort experiences the same kind of environment as theprevious one and the clear implications for the use of cohort data when making policy. For example, we need to be aware ofthe danger of basing policies for current children on evidence on children born in 1958.
Thirdly, he talked about the importance of early childhood influences (which was picked up again by Jane Waldfogel later on),emphasising how differences between children start early and get worse. This work highlights the importance of policies forearly intervention to prevent downward cycles of deprivation and create upward cycles.
Childcare and outcomes
Professor Waldfogel reviewed nine studies of the impact of childcare on future life chances. The evidence from these studieswas immensely reassuring – there are programmes that actually have an effect. She raised the interesting issue of “follow-through” – she mentioned one expansion at the top end but we’ve also had early Head Start introduced at the bottom end. Theevidence from Head Start is quite encouraging but it does raise another issue which we ought to think about ie follow through.All the evidence shows that you do need to persist if you are going to have an effect. She reviewed some of the work based onthe US National Childcare Study.
Turning to the importance of quality, it would be interesting to have a debate about what quality is. Jane Waldfogel mentionedsome of the factors. However there is a dispute between economists’ views of what constitutes quality in childcare and theprofessionals’ view of what constitutes quality. Economists tend to look at what parents want and professionals tend to look atwhat on professional grounds is good for the child. This debate has not really been entered into over here. We now have aNational Childcare Strategy but childcare for what is still a bit unclear. It would be good to make sure we don’t get swept intoa sort of “bums on seats” approach to childcare setting targets without quite asking what it is that the childcare consists of.That debate has still to be opened up.
Divorce/family breakdown
Dr Kiernan concentrated on issues concerning the extent to which divorce per se has an independent effect from all thosefactors that go along with any tendency to divorce and used both cross sectional and longitudinal data to look at the factorsaffecting divorce.
The conclusions were that allowing for financial hardship attenuates the measured impact of divorce but there still remains animpact of divorce beyond that. She then looked at the consequences of divorce. Generally demographic factors were moreclearly associated with divorce, factors such as subsequent marital breakdown, cohabiting and so on. They were rather morerobust to divorce than were some of the economic type factors. She then went on to talk about the longitudinal impacts,looking at the effects by age 33. Again the impact of the father’s interest in the child is extremely important.
It is important to look at the impact conditional on poverty; is there anything you can do and indeed is there a feedback fromwhat you can do the likelihood of being in poverty. For example, some of these programmes aren’t just affecting children butare also affecting their parents. Will any of that have an impact on the likelihood of remaining in poverty, for example aremothers likely to acquire skills in the context of a programme like Sure Start or the kinds of programmes that are already inexistence that resemble Sure Start? And will programmes to combat maternal depression, for example, have any impact ondivorce, have any impact on future participation in the labour force, and so on? These are issues that show again that whilechildren are very important and where you start from on all these sorts of programmes, the impact on parents and the two wayflow in all of this is just as important and something that we need to get right.
A number of the results Dr Kiernan showed from other work on NCDS raised the importance of the father’s interest ineducation and of fathers’ influence generally. This is an area which is very hard to come to grips with. Most childcare tendsto be discussed in terms of the mother and the needs of the mother and so on, and yet clearly programmes will make more ofan impact if they affect fathers’ behaviour as well as mothers’. However, getting dads to feel that programmes of this kind thatare about, for example, teaching them to read to their children or enabling them to read to their children or to spend time withtheir children, is really quite hard and something which we don’t know much about.
120
Session 6: Education and Poverty
‘Poverty and Education: evidence for education’s role in combating the transmission of poverty’
Ralph Tabberer, Department for Education and Employment
Introduction
Education has a crucial part to play in combatting persistent poverty. This paper has presents research evidence which supportsthis proposition. Policy ramifications will be examined in subsequent papers and discussion.
First, the paper establishes the link between education and adult life chances. Second, a question is raised about why thoseconcerned with poverty have so often been blind to education’s role. This is changing. Third, the paper explores the evidenceabout how poverty is transmitted. Not enough is known in this area.
Fourth, the evidence from school effectiveness and school improvement research is used to underline that education can makea difference. Effective approaches are identified. The paper concludes by emphasising the strategic value of the position thatschools hold in the battle with persistent poverty.
The effect of education on life chances
There is a clear relationship between education and life chances. Hobcraft (1998) demonstrates with the longitudinal NationalChild Development Study (NCDS) data that educational test outcomes are powerful predictors of adult outcomes. In fact,education is well represented among the five ‘most powerful and consistent predictors’ he identifies: childhood poverty, familydisruption, contact with the police, educational test outcomes and father’s interest in schooling.
Researchers have demonstrated that the relationship is strong. NCDS data show that, compared with those with high skilllevels, three times as many women and five times as many men with very low literacy and numeracy did not work full-timebetween 23 and 33. By the age of 37, men in the very low skills group were six times more likely to be out of work than thosemen with high skills. Studies in the United States reinforce this message. Murphy and Welch (1989) show that educationstrongly influences socio-economic status. Manski (1992) emphasises the place of a parent’s education in influencing that oftheir child.
U.S. data additionally suggest that the effect strengthens during periods of economic difficulty. Osterman (1991) shows thatwhereas in 1973, the average wage of male college graduates was 41 per cent greater than the average for high schoolgraduates, in 1991 the difference was 56 per cent. Corresponding UK data show that between 1979 and 1994/5, the net incomeof the top decile of earners – after housing costs – grew by 68 per cent, while those of the bottom tenth fell by eight per cent.
Furthermore, in deprived areas where poverty and social exclusion have been more intensely concentrated in recent years,education looks an even starker issue. A 1995 survey (Social Exclusion Unit, 1998) showed that in the secondary schoolsserving ‘difficult to let’ estates:
• one in four children gained no GCSEs;
• this is between four and five times the national average;
• truancy was four times the national average.
Education also features strongly in the report of the Social Exclusion Unit on neighbourhood renewal (SEU, 1998). The reportindicates that in the difficult estates, roughly a quarter more adults have poor literacy and numeracy skills.
The fact that on these estates, twice the nursery/primary schools and five times the secondary schools are deemed to be failing(i.e. they have been placed ‘on special measures’ after OFSTED inspection) suggests that, in too many areas, educationalprovision is unable to overcome deprivation. Indeed, it might unkindly be said that education is part of the problem rather thanpart of the solution. This is sobering, especially when it is recognised that on such estates, child density is one-fifth higher thanthe national average.
Under-expectations of education
Given the strong relationship between education and adult outcomes, perhaps it is surprising that, in this country, education’splace in anti-poverty strategies has been repeatedly under-stated. Schools are too rarely mentioned in traditional treatments ofpoverty. Textbooks address diet, health, housing, homelessness, family disruption, age, unemployment, income, ethnic origin,gender, and crime. Many key texts on poverty even fail to mention education or schools in the index.
Education is relegated to an inactive, almost invisible, role. Schools are merely places where poverty is played out. A fewauthors, of course, expressly deny education any significant position. Famously, this was the conclusion of the Coleman Report(1966). Snow et al. (1991) also argue that the effectiveness of schooling depends on the quality of the home environment.
121
More recent observers have argued that tackling economic not educational issues will raise standards.
This amounts to ‘schools-blindness’: persistent under-expectation regarding the role of schooling in helping to tackle povertyand deprivation. The answer is not to ignore education but to ascribe it a suitable and positive place.
There is evidence of changing attitudes to the importance of education, and there may be several reasons for the change:
The importance of raising productivity
In many areas of business and industry, the United Kingdom lags behind our leading competitors in terms of productivity –improved education and training will not be sufficient to raise productivity but they are now perceived as a necessary part ofthe solution.
The evidence of higher performing schools in disadvantaged areas
Benchmark tables produced annually now show the range of performance of schools in different socio-economic categories –and it is striking that, while the distribution demonstrates the overall impact of background on performance, it also indicatesthat some schools in poor areas perform about as well as the top 25 per cent of performers in the most affluent areas (seeFigure 1).
Figure 1
(Source: Excellence in SchoolsWhite Paper, July 1997)
The ‘schools make a difference’ movement
Since educators first reacted to the Coleman Report (1966), research has been pursued to demonstrate the extent to whichindividual schools make a difference to the educational outcomes of their pupils – evidence has built up about more and lesseffective schools.
The importance of integrated services
Anti-poverty strategies have increasingly recognised the need for locally-coherent, concerted action on several fronts tocombat social exclusion (OECD, 1995 and 1996) and so education increasingly comes into the frame as a possible partner.
Looking for educational solutions
To present education as the solution to the transmission of poverty, either between generations or between infancy and adultlife, would be as unhelpful as to deny education a role. And, in practice, too little is known about how poverty and lowperformance are associated in the experience of children and young people going to school.
We are not clear what makes the difference in the poor child’s day: is it whether or not they had breakfast; how well they slept;what they wear; whether or not the child has high self esteem; if they had somewhere to do their homework; if they bring toschool anxieties based on family arguments and disruption; whether or not their parents heard them read; if they weredistracted at home or in the school; or simply if they bothered to attend, or truanted? There is a strong argument for moreethnographic research, designed to reveal some of the key mechanisms for transmitting poverty into low educational outcomesand, subsequently, low outcomes, into poor adult life chances.
122
A few things are known:
• school, teacher, parent and peer expectations matter– as has been demonstrated by studies of ethnic differences (see theSwann report, 1985 and Skellington and Morris, 1992) which show how under-expectations of certain groups affectperformance;
• children’s own expectations of work matter– as has been demonstrated when young people depress their aspirationsbecause of low employment prospects (Wilson, 1987);
• some institutional barriers exist – for example, ethnic minority groups point out the bias that lies within ethnocentricassumptions about the curriculum and within national assessments;
• low basic skills at an early age can obstruct performance later– studies underline the importance of basic literacy andnumeracy for school success (Murnane, 1994);
• pupil disaffection can lead to low attendance and truancy– attitude surveys provide evidence of increasing disaffectionespecially through the first years of secondary schooling (e.g. Keys, 1994);
• family stress and emotional insecurity are factors– difficulties at home can be brought into school in the form of higheranxiety and insecurity (Venezky,1987; Snow et al, 1991);
• poor education resources at home matter– studies reveal an association between fewer books and other resources athome, and poor performance at school (Dossey et al, 1988 Snow et al., 1991);
• living in care is a major issue– there is evidence of a devastating relationship between growing up in care or beingfostered and poor educational outcomes (House of Commons, 1998);
• there is an increasing link with crime– Metropolitan Police data reveal a close relationship between youth, truancy andcrime (House of Commons, 1998);
• father’s/mother’s interest in schooling matters– recent research underlines the importance of parental attitudes toschooling (Hobcraft, 1998).
Despite these findings, we are not clear which are the antecedents and which are the consequences, let alone which are thecauses and which are the effects.
And at this point, we must be careful to draw upon the evidence about persistent poverty. It is not simply the case that areadily-defined group, or under-class, abides in poverty, transmitting their culture and experience from one generation to thenext.
Bradshaw and Holmes (1989) studied poor families in [North Tyneside] and concluded:
“… it is of crucial importance to recognise that these families, and probably many millions more like them living on socialsecurity benefits, are in no sense a detached and isolated group cut off from the rest of society. They are just the same people asthe rest of our population, with the same culture and aspirations but with simply too little money to be able to share in theactivities and possessions of everyday life with the rest of the population.”
This may be over-stating the case but, for many, the experience of poverty is indeed intermittent. Regularly, families emergefrom poverty, some to escape and others to fall back quite quickly, perhaps due to a new bout of unemployment or the birth ofa new child. We need to understand the impact of temporary as well as recurrent poverty on educational experiences andoutcomes, before we can best judge how to change education to improve the prospects for specific individuals and groups.
Schools make a difference
While too little is known about how poverty and failure are experienced or related at school, there is a strong body ofknowledge about what contributes to effective schooling. Over twenty years, a considerable body of evidence has beengenerated to support the contention that a school can make a real difference to the educational achievement of its intake.
There are many studies, most of them in North America but with several significant studies in the UK, the Netherlands andAustralia as well. There are many research reviews.
For instance, the main studies include:
• Brookover and Lezotte (1977) who provided early evidence of the relative effectiveness of eight primary schools inMichigan, USA;
• Rutter et al.(1979) who provided evidence of school effectiveness in 12 secondary schools in this country;
123
• Mortimore et al. (1988) who covered 49 primary schools;• Tizard, B. et al.(1988) who addressed effectiveness in infant schools; and
• Smith and Tomlinson (1989) who examined effectiveness in relation to ethnic minorities.
The effectiveness studies put the size of the school effect at about eight to ten per cent. There is some debate about this figureand about what it means. Gray et al.(1990) found the difference between schools at the upper as opposed to the lower quartile,to be around four points per pupil at GCSE, and that has been described as giving young people a ‘competitive edge’. Therange of difference between the most and least effective schools was, of course, much wider. Reynolds et al. (1996) points outthat the size of the school effect is much greater in countries like the USA and UK than it is on the Pacific Rim.
The studies have helped clarify those factors which are associated with more effective schools, where effectiveness is primarilymeasured on educational test scores. The conventional list of key factors is provided by Mortimore (1991):
Leadership
A leader who is purposeful, neither too authoritarian nor too democratic, who is able to share ownership of the school withcolleagues, who can delegate to a deputy without feeling threatened, and who can involve staff in planning and management.
Management of pupils
Pupils are involved, they can be rewarded for effort; control of behaviour is by methods neither too weak nor too harsh;teaching sessions are structured, work-centred and include intellectually challenging teaching.
Management of teachers
Teachers are involved in the corporate life of the school; they pursue consistency in their approach to pupils, and areencouraged to be good models of punctuality, politeness and consideration; classrooms have positive psychological climates inwhich pupils are encouraged to communicate frequently with teachers; there is a broad and balanced curriculum whichrecognises the academic but also values the special needs students – and in primary, there is a limited focus in teachingsessions to avoid teachers and students being pulled in different directions.
Pupil care
Pupils are treated with dignity and encouraged to participate in the organisation of the school; there are positive signals thatpupil are valued; rewards are used rather than punishment to change behaviour; parents are involved in the life of the schooland there is increasing confidence of the community in the school; systematic records of pupil progress enable the curriculumto have coherence for pupils.
School environment
Attractive and stimulating, with trouble taken over class displays and removing graffiti.
School climate
Consensus is sought on values shared by the school, the general attitude towards learning is positive as is the attitude to youngpeople; there are clear rules and guidelines for pupil behaviour; high expectations are maintained for all pupils.
In its study of schools which were successful in disadvantaged areas, achieving good results ‘against the odds’, the NationalCommission on Education (1995) emphasised the importance of:
• leadership fostering a team approach;
• a vision of success, and pride in achievement;
• policies and practices which encourage planning and the setting of targets;
• improvement of the physical environment;
• common expectations about pupil behaviour and success;
• investment in good parental and community relationships.
School improvement programmes have built on the school effectiveness research, especially in the USA (see Comer, 1980;Levin, 1988; Sizer, 1992; and Slavin 1989). These school-based reforms attempt to change the daily experience children haveof school. Indicative of the approach is that taken by Slavin under the ‘Success for All’ programme. This emphasises: an earlyyears focus, a literacy core including a literacy instruction hour, proven instructional and cooperative learning techniques,regular assessments which lead to immediate or early recovery for those failing, a parental support programme, and appropriateteacher training.
124
Schools taking part need to have an 80 per cent majority of teachers in favour of adopting the approach. The adoption of eachelement in the approach is based on research evidence about what works, and Slavin underlines the importance of ‘highfidelity’ implementation, and relentlessness.
These programmes have informed system-wide reforms around the world: in California, Kentucky, Texas and North Carolinain the USA; in Victoria, Australia; and now in the UK. Again, the logic is that no single measure will change the dailyexperience of children; only the sustained implementation of several measures will be enough.
Knowing education has an important role is not enough
Based on his primary school researches, Mortimore (1997) points out the possibility and the limitations of education’s role:
“… in no school did we find evidence of whole groups of the disadvantaged outperforming whole groups of the advantaged.We did find, however, that disadvantaged students in the most effective schools sometimes made more progress thanrelatively advantaged students in the least effective schools”.
The challenge for education, if it is to be a successful part of an anti-poverty strategy is to find out which improvementprogrammes achieve the twin goals: of raising standards and closing the gap between top and bottom.
Understandably, this is difficult. In a report based on the first four years of secondary school inspection in England, OFSTED(1998) reported that although the quality of management and leadership improved between 1993 and 1997, the quality ofteaching improved, and standards of achievement were raised, the gap between high performing and low performing schoolswidened.
The simple truth is that those who start with an advantage are usually better placed to exploit any new system or initiative.There is, however, some evidence about the educational measures which favour those who come from disadvantagedbackgrounds. For example, Murnane (1994) identifies initiatives which have particularly worked to improve standards amongthe poor in the USA, including:
• pre-school programmes for disadvantaged children based on Head Start (Barnett, 1992);
• vocational training which is integrated with traditional subjects in order to provide better preparation for work (Resnick,1987);
• greater incentives to teachers in urban districts, e.g. providing college grants in return for service in disadvantaged areas(Arfin, 1986).
There are more. For example:
• the Project STAR class size experiment in Tennessee revealed that markedly lower class sizes in the early years led togreater gains for pupils from disadvantaged areas (Nye et al., 1993; Achilles et al., 1993);
• the ‘Success for All’ programme (Slavin, 1990) has delivered some of its highest gains where pupils enter school with poorlanguage acquisition;
• Edmonds (1989) reports that an emphasis on basic skills instruction (reading and mathematics) favours more disadvantagedpupils;
• two-generation family literacy programmes have been effective both in the USA and the UK (see Brooks et al., 1996).
These research findings support many elements in the Government’s current strategy for raising standards and closing the gapbetween low and high performers. The task is now to consider what further action might be appropriate to achieve even greaterleverage.
Schooling is a key life experience
15,000 hours of schooling is more than a passage in the life of the poor, an unimportant sequence of events somewherebetween childhood poverty and diminished adult life chances.
Schools are significant institutions in their localities and represent a valuable capital and human resource investment for theircommunities. In the most difficult areas, starkly, they may even be the only moral community to which young people everbelong. The State has few mechanisms which have as a matter of course so much access to the young and poor.
The experience of school can be neutral: a time when inherited poverty is transmitted into low skills and high risks of reducedadult life chances. The experience of schooling can even be negative: a time when disadvantages are exacerbated by even moredisadvantaged educational opportunities. Or the experience can be to some extent positive and, with other measures,compensatory.
125
It is wise to consider education as part of the solution. It may even be wise to start to think of schools as a more important basefor the combined, inter-agency and community solutions that will be needed to improve cultural capital in deprived areas, tocombat local poverty and to tackle social exclusion.
This is not to argue that schools should be diluted. The teaching function must remain targeted at educational test outcomes.Teachers should not be distracted and try to become social workers, youth workers, counsellors, mentors, care workers andadult literacy tutors.
But the school’s work in raising standards may be enhanced by being more closely located alongside other communityresources. Then, integrated and localised services may be able both to raise standards and combat the transmission of povertyfrom childhood to adult life.
References
Achilles, C.M. et al. (1993). ‘The Lasting Benefits Study (LBS) in grades 4 and 5 (1990-1991): A Legacy from Tennessee’s Four Year (K-3) Class Size Study(1985-89).’ Paper to North Carolina Association for Research in Education.Arfin, D.M. (1986). ‘The use of financial aid to attract talented students to teaching: lessons from other fields’, Elementary School Journal, 86, p405-23.Barnett, W.S. (1992). ‘Benefits of compensatory preschool education’, Journal of Human Resources, 21 (Winter), p1-23.Bradshaw, J. and Holmes, H. (1989). Living on the Edge: a Study of the Living Standards of Families in Benefit in Tyne and Wear. Tyneside: CPAG.Brooks, G., Gorman, T., Harman, J., Hutchison, D. and Wilkin, A. (1996). Family Literacy Works. London: Basic Skills Agency.Brookover, W. and Lezotte, L. (1977). Changes in School Characteristics Co-incident with Changes in Student Achievement. East Lansing Institute forResearch on Teaching, Michigan State University.Coleman, J.S. et al. (1966). Equality of Educational Opportunity. Washington DC: Department of Health, Education and Welfare.Comer, J.P. (1980). School Power: Implications of an Intervention Project. New York: Free Press.Dossey, J.A. et al. (1988). The Mathematics Report Card: Are We Measuring Up? Report 17-M-01. Princeton, NJ: Educational Testing Service.Edmonds, R. (1989). ‘Effective schools for the urban poor’ in B. Cosin, M. Flude and M. Hales (eds.). School, Work and Equality. London: Open UniversityPress.Hobcraft, J. (1998). Intergenerational and Life-Course Transmission of Social Exclusion: Influences of Childhood Poverty, Family Disruption, and Contactwith the Police. Centre for Analysis of Social Exclusion Casepaper 15. London: LSE.Gray, J., Jesson, D. and Sime, D. (1990). ‘Estimating differences in the examination performances of secondary schools in six LEAs: a multi-level approach toschool effectiveness’, Oxford Review of Education, 16, 2, p137-58.Keys, W., Harris, S. and Fernandes, C. (1995). Attitudes to School, Slough: NFER.Levin, H.M. (1988). Accelerated schools for at-risk students. Centre for Policy Research in Education. Report Series RR-010. Ruttgers Univ, Brunswick, NJ.Manski, C.F. et al. (1992). ‘Alternative estimates of the effects of family structure during childhood on High School graduation’, Journal of the AmericanStatistical Association, 87, p25-37.Mortimore, P. et al. (1988). School Matters: the Junior Years. Salisbury: Open Books.Mortimore, P. (1991). ‘The nature and findings of research on school effectiveness in the primary sector’. In: Riddell, S. and Brown, S. (eds.) SchoolEffectiveness Research: Its Messages for School Improvement. Edinburgh: HMSO.Mortimore, P. et al. (1997). ‘Can effective schools compensate for society?’ in A.H. Halsey et al. (eds.). Education: Culture, Economy and Society. Oxford:Oxford University Press.Murnane R.J. (1994) ‘Education and the well-being of the next generation’, in S.H. Danziger, G.D. Sandefur and D.H. Weinburg (eds.). Confronting Poverty.Prescriptions for Change. London: Harvard University Press.Murphy, K.M. and Welch, F. (1989). ‘Wage premiums for college graduates: recent growth and possible explanations’, Educational Researcher, 18 (4), p17-26.National Commission on Education (1995). Success Against the Odds: Effective Schools in Disadvantaged Areas. London: Routledge.Nye, B., et al. (1993). ‘Tennessee’s bold experiment’, Tennessee Education, 22/3, p10-17.OECD. Centre for Educational Research and Evaluation (1995). Our Children at Risk. Paris: OECD.OECD. Centre for Educational Research and Evaluation (1996). Successful Services for our Children and Families at Risk. Paris: OECD.OFSTED (1998). Secondary Education 1993-97. A Review of Secondary Schools in England. London: OFSTED.Osterman, P. (1991). ‘Is there a problem with the youth labor market; and if so, how should we fix it?’. Sloan School of Management. Massachusetts Instituteof Technology. Mimeo.Reynolds, D., Creemers, B., Stringfield, S. and Teddlie, C. (1996). ‘World class schools: some further findings.’ Paper presented at AERA annual conference.New York.Sizer, T.R. (1992). Horace’s School: Redesigning the American High School. Boston: Houghton Mifflin.Skellington, R. and Morris, P. (1992). ‘Race’ in Britain Today. London: Sage.Slavin, R.E., Karweit, N.L. and Madden, N.A. (1989). Effective Programs for Students at Risk. Boston: Allyn and Bacon.Slavin, R.E.(1990). ‘Class size and student achievement: is smaller better?’, Contemporary Education, 62/1, p6-12.Smith, D. and Tomlinson, S. (1989). The School Effect: A Study of Multi-Racial Comprehensives. London: Policy Studies Institute.Snow, C.E. et al. (1991). Unfulfilled Expectations: Home and School Influences on Literacy. Cambridge, Mass: Harvard University Press.Social Exclusion Unit (1998). Bringing Britain Together: a National Strategy for Neighbourhood Renewal. Cm 4045. London: HMSO.Swann Report (1985). Education for All: the Report of the Committee of Enquiry into the Education of Children from Ethnic Minority Groups. London:HMSO.Tizard, B. et al. (1988). Young Children at School in the Inner City. London: Lawrence Erlbaum Associates.Venezky, R.L., Kaestle, C.F. and Sum, A.M. (1987). The Subtle Danger. Reflections on the Literacy Abilities of America’s Young Adults. Princeton, NJ:Educational Testing Service.Wilson, W.J. (1987). The Truly Disadvantaged: The Inner City, the Underclass and Public Policy. Chicago: University of Chicago Press.
126
‘Childhood poverty and post-compulsory participation and attainment. causal effect and impact onlifetime inequality’
David Soskice, No. 10 Policy Unit
Introduction
I want to change the tone of the discussion from this interesting and persuasive statistical discussion, to asking some questions– which actually relate closely to the previous discussion about the way in which incentives work for these groups of youngpeople. I will do this by looking at the difference between UK experience and German experience to show how an environmentvery different to the UK creates reinforcing incentive structures for young people, for teachers, for parents. By incentivestructures, 1 am not referring to short-term financial incentives (though they may have a part to play), but rather in the Germancase to the clarity of the links from educational performance to acceptance in apprenticeships and from performance inapprenticeships to well-rewarded employment. It is this linkage, and the fact that it is generally understood by students, parentsand teachers, which leads to students of all ability levels taking education seriously. Some people say that the Germans areculturally predisposed to take education seriously: that is true but superficial. Culture reflects and is reinforced by embeddedincentives; without those incentive structures, the culture quickly collapses.
The purpose of looking at the German case is to understand the importance of clear incentive structures linking education toemployment. I will not argue that we should try and copy the Germans and build a major apprenticeship system in the UK: thesecond purpose of looking at Germany is to show that the institutions which underpin their system and the way their labourmarkets work cannot be copied in here. Rather, that our task is to build incentive structures for the bottom third of youngpeople who cannot see credible links between education and good employment prospects.
The German system
I’m going to first talk about the way the German system works and then about the British system. Everything of course comesdown to the relationship of the perception which young people have, of the relationship between taking school seriously andsubsequent employment. In the German system there is a very clear understanding which almost all the participants have of therelationship between working hard in school and first training and then employment. How does that work? As I’m sure most ofyou know about the German system, about 60% of young people at the age of 16 or 18 do 31/2 year apprenticeships andapprenticeships more or less go down to the bottom 6 or 7% of the German age cohort. There are a group of children at thebottom who are largely from immigrant backgrounds who drop out of the system. Critically for incentive structures, there is avery big spectrum of apprenticeships: they go from very highly esteemed apprenticeships in major companies, the major banks,Daimler-Benz and so on down to low level apprenticeships in the so called hand work sector, the small company sectorincluding the retail sector, garage mechanics and so on. Within any locality, parents, teachers and hence children have prettygood idea of this distribution of apprenticeships. There is moreover usually a close connection usually between companies andschools. Companies often have long term relations with local schools: they have therefore a good idea from teachers aboutwhich children work hard at school, as well as exam results. There is therefore an incentive structure for young Germans inrelation to apprenticeships which is almost exactly the same incentive structure which we have in this country, not at thebottom or middle end of the educational process but at the top end when it comes to scoring A Level points to get intouniversities. In the UK, the better the A-level grades that you get the better the University you can go to. In Germany, at thebottom and middle end of the system there is exactly the same incentive structure: the better you do at school, the better you’reperceived to do by your teachers and the better the exam results you get at 16 and 18, the better the apprenticeship which youcan get into. Thus even young people right at the bottom of the achievement ladder in German schools have got an incentive todo that much better and improve the apprenticeship which they get. (In Germany as it happens there is no such incentivestructure at the top end as far as university entrance is concerned, you merely have to pass the Arbitur; with some exceptionsthe level of pass does not matter.) This means that teachers, children and parents are all highly aware from a pretty early age ofthe importance which attaches to teachers being persuaded that children take school seriously and that they get good examresults.
Now this sounds an ideal system and, of course, people have tried on many occasions to develop such a system in this country.The problem is that the apprenticeship system in Germany rests on important institutions, outside the education system andmore or less outside the apprenticeship system, which enable the apprenticeship system to work well. The importance of theseinstitutions can be seen in the answers to the following questions:
• Why do young people have confidence in the value of apprenticeships in Germany?
• Why are companies in Germany prepared to invest heavily in apprenticeships?
Young people have confidence in the value of apprenticeships because it is fully understood that the standards of trainingwhich you learn in your apprenticeship will be of use across a wide range of employers within the occupation. That’s becausethere are effective methods for employers to reach consensus on what the key training vocational standards are which theywant their own employees to use. Behind that lies the fact that powerfully resourced employer associations can bring aboutconsensus among employers, something which it is very difficult to imagine having in the UK.
A second reason why a young person has confidence in the value of an apprenticeship lies in the high regulation of German
127
labour markets and in particular the stability of German collective bargaining. If you go into one occupation you have a prettygood idea that the earnings you’ll get in that occupation won’t move dramatically down compared to other occupations, so youhave a guarantee from the way in which the collective bargaining system works in Germany, that by investing in anapprenticeship you’re investing in something which will insure that your wages more or less rise in relation to wages elsewherein the economy. This very stable distribution of wages over time in the German economy depends on powerful unionsoperating with powerful employer associations. Again, there is no way in which such a development could take place in themodern British economy.
Thirdly there is relatively limited occupational mobility; people don’t move around as they do in this country across a wholenumber of occupations, so that the value of the vocational training which they have is one which remains valid for a muchlonger period of time. Finally the training takes place half in companies, half in state vocational schools (dual system). Withinthe company strong employee elected works councils monitor the quality of the training which a young person is getting.
Why are companies prepared to invest in apprenticeships? The first reason is interesting because it contrasts very much withthe UK. German companies know that they will be able to train young people relatively easily in complex skills since theapprentices they take on start off with a good basic educational level. Young people on the whole have reasonable literacy andnumeracy capabilities before they start work related training. Secondly, companies in Germany have much less to fear from thepoaching of skilled labour which takes place in this country; this results from regulated labour markets and again this comesback to powerful employer associations and unions. Thirdly, companies have access to long term finance in Germany and thatmeans that they’re prepared to take the relatively long term investment decisions which investing in apprenticeships implieswithout the immediate worry of their bottom profit line.
What can we learn from Germany?
In Germany institutions play a major part in constructing and supporting what might be described as a positive feed-backincentive structure for young people, as well as for their teachers and parents. This shows the value, in terms of subsequentemployment, of taking education seriously; and it applies even to most children from low income backgrounds and theirteachers. Because we do not have (and probably would not want) Gennan-type institutions, we should not think in terms ofimporting German solutions. What we can learn from Germany is the importance of effective incentive structures. We need totry and understand why the incentive structure for the bottom third of young people and their teachers in the UK is weak andappears to operate in a counter-productive way.
In particular, there is a lot of evidence which suggests that those who stay on between 16 and 18 or 19 in full-time education orhighly structured training can make the transition into good employment. What is it about the incentive structures facing youngpeople and/or their teachers or educational institutions which leads the bottom third of young people to leave education at 16,frequently with low or no GCSEs?
First, these most of these young people have grown up in an environment in which, if jobs are available at all, they are eitherunskilled or semi-skilled. The great likelihood is therefore that they will end up in this end of the labour market. Here the needfor literacy and numeracy, and academic achievement in general, is much less great than it is in the German system. Typicallyof the bottom third only about a third of sixteen year old school leavers will actually get employment directly. Many will moveinto unemployment, or move out of the labour market, and some will move into Government training schemes. Its very clearthat there is no requirement that you work hard at school in order to qualify for such career moves.
Secondly, if one looks at the incentive structures for schools up to the age of 16 (when most of these young people leave theeducation system), they have a very limited incentive to devote resources to these young people. The flindamental incentivestructure for most secondary schools and many primary schools is to recruit middle class children. In order to do so, thesecondary school needs to show it is successful in GCSE A to C passes. Therefore it makes little sense for a school to devoteresources to moving children up from E’s or F’s to D’s . It makes sense for the school to devote resources moving D’s up to C’sand to moving other children up higher.
Third, the same lack of incentive to invest in these children is true for school sixth forms, sixth form colleges and FE colleges.Trying to attract children who are low achievers at the age of 16, and devoting resources to literacy and numeracy and personalcounselling is costly. There are different incentive structures for school sixth forms and for FE colleges, but they have similarimplications. In general the incentive structure for school sixth forms is to devote resources to A Level children; their successdepends upon attracting middle class children. The incentive structure for FE colleges is to pack as many children in aspossible but ensuring that these are children who can pass courses so that the college can get its full payment.
Fourth, young people with low/no GCSEs at the age of 16 have very limited incentives to stay on in FE colleges or schoolsixth forms: If they can get employment they can earn money; if they move into work related training in Governmentsupported training they can also earn money. But they receive nothing if they stay in full-time education. Hence they have littleincentive to stay on at school or college in the post-16 phase if they are low GCSE achievers. This is reinforced by the attitudeof schools and colleges.
Low literacy and numeracy makes difficult the move from insecure low-level labour markets in which little training takes placeinto more satisfactory employment. Moreover this low level employment environment, interspersed with unemployment orperiods outside the labour market altogether, is not one in which social skills become developed. In an increasingly service
128
sector economy, social skills become ever more important. But social skills demand that young people grow up in a mixedsocial environment; and that is the critical role of education between 16 and 18 or 19 in the same institution – albeit doingdifferent courses.
Until we can correct incentive structures for young people and educational institutions we are unlikely to be able raise stayingon rates and achievement substantially for initial low achievers. It cannot be done be changing the nature of employment –unskilled low level jobs will only decline as the number of low achievers without social skills declines. We need therefore tooperate on the incentives which face colleges and schools as well as the financial incentives for young people between 16and 19.
Finally to underline this reinforced circle of low incentives in the UK companies have very limited incentives indeed ininvesting in children who are low achievers at the age of 16, the strong incentive for German companies that what they getfrom schools are young people even at a low level who have got good literacy and numeracy abilities and have shownthemselves to be responsible and the like, companies simply do not have in this country in thinking about training and henceupgrading these sorts of children so you have a whole set of incentive structures all of which operate in the wrong directionand I feel rather strongly that this is in a way much broader and sketchier way than the previous speaker reinforces the pointswhich he was making.
129
130
Session 6: Education and poverty: Discussion
Table 1 shows the obvious links between earnings and education.
Table 1: Crude Earnings by Highest Qualification (all ages)
Source: Spring 1998 Labour Force Survey.
Education represents a critical part of the strategy to raise opportunity for all. But for any given intake of pupils (crudelymeasured by eligibility for free school meals) schools are not yet delivering the higher attainment for all which will have asignificant impact on levels of poverty. How could the educational process be made more reliable ?
Second, how equal could opportunity be in practice across different school intakes? This in turn leads one to ask whether weneed a distribution of resources even more skewed towards those schools or pupils who are not achieving.
In order to make the progress required in education, there needs to be complementary policies for family support to mitigatethe impact on children of low educational attainment by their own parents or financial difficulties in the family, which Greggand Machin1 show has a major impact on the chances of young adults staying on in education.
Assuming these were in place, the potential benefits of education include:
• attainment at any stage of education mapping on to higher earnings (as indicated in the table comparing differentattainments at GCSE);
• staying-on post 16 did likewise;
• socialisation and preparation for citizenship probably contributed to this; as did
• a greater likelihood, as attainment rose, that employers would spend more on an individual’s training; and
• a likelihood of delay in starting a family.
Given the contribution education could make to earnings, it would help individuals to find the likeliest exit routes frompoverty. Stephen Jenkins’ analysis indicates that 63% of such exits are associated with changes in earnings.
The price of failure in education and associated family support policies included inter-generational transmission of poverty:Gregg and Machin show that children aged 6-9 of those in the NCDS study had lower test scores than average if one or both oftheir parents had had low school attendance, if the parent’s family had suffered financial difficulty when s/he was a child, ifs/he had been in contact with the police as a child or if s/he had ever been in care.
So are there three questions for policy:
i. We have the intention to level up standards; Ralph Tabberer’s presentation had suggested that we have theknowledge. Do we have the leverage?
ii. How equal can opportunity be – do we have the methods and incentive structures to equalise?
iii. Are the complementary family support policies in place (or else even more interventionist schools, such as thosenow providing breakfast for their children)? The New Deal and the Working Families Tax Credit would contributehere.
Education is likely to be even more important in the future than it has been to date with the decline of manual employment andthe diminution of other sources of comparative national advantage; we can measure a dispersion of outcome and seem able toidentify necessary conditions for success; so we could choose priorities and set targets.
£ per week None 1-4 GCSEs 5+ GCSEs 2+ A levels First degreegrade A-C grade A-C
Men 262 307 371 438 554
Difference fromprevious column 17% 21% 18% 26%
Women 127 174 210 256 345
Difference fromprevious column 37% 21% 22% 35%
1 Child Development and Success or Failure in the Youth Labour Market (Centre for Economic Performance discussion paper 397, 1998).
Discussion
A number of questions were raised from the floor.
• How do we create a virtuous circle of good education and good jobs, or break out of the vicious circle of ‘steppingstone’ jobs which perpetually lead to other ‘stepping stone’ jobs?
David Soskice identified staying on rates as crucial. He highlighted the importance of changing the incentive structure for 16-19s. One important vehicle for this would be the new Education Maintenance Allowance to be piloted later this year, whichshould be performance related.
Ralph Tabberer suggested addressing the problem using merit pay to attract the best and brightest to teaching, and rewardsuccess achieved despite difficult circumstances. This point was reinforced, citing the improvements achieved in Dallas, Texasthrough such schemes. The recent introduction of a programme allocating increased funding to FE colleges for recruiting fromspecific postcodes of areas of high deprivation was also referred to.
• Are test scores appropriate measures of achievement – they may just reflect IQ? Is education as effective as issuggested, developing children rather than simply ‘jumping through hoops’?
The participants pointed out the importance of focussing on results as the best available indicator of performance.
• In response to David Soskice’s presentation, how is an appropriate political balance established with employers inGermany, particularly on issues such as curriculum control?
David Soskice had emphasised the stake that employers had in ensuring the effective working of the system of apprenticeships,and the need of all parties involved to establish a balanced and effective working relationship. He went on to discuss theproblems caused in the UK by labour market de-regulation and casualisation of employment. This was leading to a perceptionof employment as a less secure option than other activities, like crime.
131
132
CASE Publications
CASEpapers,a series of discussion papers produced by the ESRC Research Centre for Analysis of Social Exclusion at theLondon School of Economics, are available free of charge, as is our series of CASEbriefs, which summarise the research.Particular conferences and activities are summarised in our occasional CASEreports series.
All these are available either from Jane Dickson, the CASE Administrator (tel: 0171 955 6679; fax: 0171 242 2357;email: [email protected]) or can be downloaded from our internet site: http://sticerd.lse.ac.uk/Case.
CASEpapers
CASE/1 Kathleen Kiernan The Legacy of Parental Divorce: Social, economic anddemographic experiences in adulthood
CASE/2 Tania Burchardt Boundaries between Public and Private Welfare: A typology andmap of services
(more detailed report, Private Welfare and Public Policy, by Tania Burchardt, John Hills and Carol Propper,prepared for the Joseph Rowntree Foundation, January 1999)
CASE/3 Jane Waldfogel, Maternity Leave Policies and Women’s Employment afterChildbirth: Yoshio Higuchi and Masahiro Abe Evidence from the United States, Britain and Japan
CASE/4 A B Atkinson and John Hills (editors) Exclusion, Employment and Opportunity
CASE/5 Julian Le Grand and Phil Agulnik Tax Relief and Partnership Pensions
(version published in Fiscal Studies, vol.19, no.4, 1998)
CASE/6 Simon M Burgess and Carol Propper Early Health Related Behaviours and their Impact on Later LifeChances: Evidence from the US
CASE/7 Kathleen Kiernan and Ganka Mueller The Divorced and Who Divorces?
CASE/8 Christian Schluter Income Dynamics in Germany, the USA and the UK: Evidencefrom panel data
CASE/9 Simon M Burgess and Carol Propper An Economic Model of Household Income Dynamics, with anApplication to Poverty Dynamics among American Women
CASE/10 Didier Jacobs Social Welfare Systems in East Asia: A Comparative AnalysisIncluding Private Welfare
CASE/11 Mark Kleinman Include Me Out? The new politics of place and poverty
CASE/12 Brian Barry Social Exclusion, Social Isolation and the Distribution of Income
CASE/13 John Hills Thatcherism, New Labour and the Welfare State
CASE/14 Irwin Garfinkel, Sara McLanahan, Fathers under Fire: The revolution in child support enforcement inDaniel Meyer and Judith Seltzer the USA
(This CASEpaper is a summary of the book by the same title and authors, published by the Russell SageFoundation, 1998)
CENTRE FOR ANALYSIS OF SOCIAL EXCLUSION
An ESRC Research Centre
ASE
CASE/15 John Hobcraft Intergenerational and Life-Course Transmission of SocialExclusion: Influences of Childhood Poverty, Family Disruption,and Contact with the Police
CASE/16 Frank A. Cowell and Christian SchluterMeasuring Income Mobility with Dirty Data
CASE/17 William Julius Wilson When Work Disappears: New Implications for Race and UrbanPoverty in the Global Economy
(to be published in Ethnic and Racial Studies, 22(3), May 1999)
CASE/18 Ross Hendry Fair Shares for All? The development of needs based governmentfunding in education, health and housing
CASE/19 Nigel Campbell The Decline of Employment Among Older People in Britain
CASE/20 Jane Falkingham Welfare in Transition: Trends in poverty and well-being in CentralAsia
CASE/21 Jane Waldfogel Early Childhood Interventions and Outcomes
CASE/22 Howard Glennerster, Ruth Lupton, Poverty, Social Exclusion and Neighbourhood: Studying the areaPhilip Noden and Anne Power bases of social exclusion
CASEbriefs
CASEbrief 1 Boundaries between public and private welfareCASEbrief 2 Maternity leave policies and women’s employment after childbirthCASEbrief 3 Exclusion, employment and opportunityCASEbrief 4 Tax relief and partnership pensionsCASEbrief 5 The State of WelfareCASEbrief 6 Who are the divorced and who divorces?CASEbrief 7 Social Welfare in East Asia: Low public spending but low income inequality?CASEbrief 8 Childhood experiences and the risks of social exclusion in adulthoodCASEbrief 9 The Decline of Employment Among Older People in Britain
CASEreports
CASEreport 1 Anthony Lee and John Hills New Cycles of Disadvantage? Report of a conference organised byCASE on behalf of ESRC for HM Treasury
CASEreport 2 John Hills (editor) CASE Annual Report 1997/98
CASEreport 3 William Julius Wilson, Geoff Mulgan, Welfare Reform: Learning from American Mistakes? Report of aJohn Hills and David Piachaud seminar organised by LSE Housing and CASE
CASEreport 4 Liz Richardson Tackling Difficult Estates: CASE/Social Exclusion Unit Seminar
CASEreport 5 Persistent Poverty and Lifetime Inequality: The evidence. Reportof a seminar organised by HM Treasury and CASE)
133
Table 1. Time spent in poverty over a six-year perioda
Share of individuals in poverty for 1 to 5+ yearsb (%)Post-tax and transfers
1 year 2 years 3 years 4 years 5+ years Average number of years in povertyCanada 1990-95 35.9 27.0 14.4 9.2 13.5 2.4Germany 1991-96 45.6 19.4 12.0 7.6 15.5 2.4United Kingdom 1991-96 26.0 19.3 13.6 13.2 27.9 3.1United States 1989-93 33.0 18.5 11.2 10.1 27.3 3.0
Share of total time spent in povertyc (%)Post-tax and transfers
1 year 2 years 3 years 4 years 5+ yearsCanada 1990-95 14.8 22.1 17.7 15.2 30.3 –Germany 1991-96 19.2 16.3 15.2 12.8 36.4 –United Kingdom 1991-96 8.3 12.3 13.0 16.9 49.6 –United States 1989-93 11.1 12.4 11.2 13.5 51.8 –
a) The sample used includes all those individuals interviewed in each of the six years who have experienced at least one year in poverty.b) The share of individuals spending one to five+ years in poverty as a share of all individuals touched by poverty over the period. For example, 35.9 per
cent of those poor during the six-year period in Canada were poor for one year and 13.5 per cent for five years or more.c) The following steps were used to calculate the values in each column. First, the values in each of the columns of the top panel were multiplied (weighted)
by the number of years spent in poverty shown in the heading (distinguishing between five years and six+ years). A weight of six was given to groupswhich have six or more years in poverty, thus biasing downward the last column in the bottom panel. Second, these weighted values were then summedto estimate the total number of years spent in poverty by the total population. Finally, the values in the columns of the bottom panel are the results of thefirst step divided by the total calculated in the second step.
Source: OECD Secretariat.
Table 2. Poverty dynamics: “hazard rates”a
Probability of exiting poverty afterb: (rate *100)Post-tax and transfers
1 year 2 years 3 years 4 yearsCanada 1990-95 55.7 41.3 38.8 35.4Germany 1991-96 52.7 42.7 32.0 19.1United Kingdom 1991-96 45.4 37.0 32.3 25.8United States 1989-93 45.6 31.9 23.1 20.2
Probability of re-entering poverty afterc: (rate *100)Post-tax and transfers
1 year 2 years 3 years 4 yearsCanada 1990-95 16.7 9.7 7.9 7.1Germany 1991-96 25.6 13.0 17.5 15.5United Kingdom 1991-96 32.8 18.2 11.0 10.0United States 1989-93 31.8 21.5 18.3 18.6
a) Latest available six-year period. Includes all poverty spells or spells above the threshold where the beginning of the spell can be observed.b) This is calculated as the ratio of those individuals, having just fallen below the threshold, who exit before the end of one, two, three or four years in
poverty relative to those still poor at the beginning of each successive period. For example, in Canada, 55.7 per cent leave before the end of the first yearand, of the 44.3 per cent who remain, 41.3 per cent leave between the first and second year.
c) The sample includes all those spells above the poverty threshold, conditional on the individual having exited poverty immediately before. The re-entryhazard is calculated as the ratio of those individuals observed to fall back below the threshold before one, two, three or four years above the poverty lineover the population at risk. For example, of those who moved above the threshold and are still above the poverty line after one year, 9.7 per cent will fallback below the threshold in Canada between first and second years.
Source: OECD Secretariat.
1
Table 3. Characteristics of individuals above and below the poverty thresholdPer cent share of persons with a specified characteristic in each group
Household characteristics Total Above Poor for Always populationa thresholda 1 yearb poora
CANADAHead genderHead male 83.7 87.9 70.6 72.4Head female 16.3 12.1 29.4 27.6Employment statusNo worker 18.3 13.6 45.3 36.4One worker 31.2 27.9 42.3 43.9Two workers 39.0 44.3 11.3 17.9More than two workers 11.5 14.2 1.1 1.8Family typeSingle adult, no children 19.4 16.1 34.3 24.9Two adults, no children 30.1 32.4 15.5 26.7Single adult, children 4.4 2.1 11.3 11.6Two adults, children 31.5 32.4 33.2 29.2Large families 14.6 16.9 5.7 7.6Age of household headc
Young-age head 28.3 25.0 38.2 28.6Prime-age head 34.0 35.7 34.8 33.4Older-working-age head 21.8 22.6 21.6 22.0Retirement-age head 15.9 16.7 5.4 16.0Education leveld
Low education – – – –Middle education – – – –Higher education – – – –
GERMANYHead genderHead male 75.2 79.1 53.1 16.7Head female 24.8 20.9 46.9 83.3Employment statusNo worker 18.5 14.9 49.4 71.8One worker 39.4 37.6 44.1 28.2Two workers 34.8 39.4 2.5 0.0More than two workers 7.3 8.1 4.0 0.0Family typeSingle adult, no children 14.4 12.3 27.6 34.2Two adults, no children 40.9 43.2 31.7 14.9Single adult, children 2.8 1.3 15.5 32.1Two adults, children 33.6 34.7 19.0 16.7Large families 8.4 8.4 6.2 2.2Age of household headc
Young-age head 12.8 10.1 24.9 31.0Prime-age head 45.8 47.8 40.4 31.5Older-working-age head 26.6 27.7 17.2 10.7Retirement-age head 14.8 14.4 17.5 26.8Education leveld
Low education 28.0 26.0 29.6 30.5Middle education 52.3 52.7 60.9 63.0Higher education 19.7 21.4 9.5 6.6
Table 3 (cont’d). Characteristics of individuals above and below the poverty thresholdPer cent share of persons with a specified characteristic in each group
Household characteristics Total Above Poor for Always populationa thresholda 1 yearb poora
UNITED KINGDOMHead genderHead male 67.7 74.3 53.7 38.4Head female 32.3 25.8 46.3 61.6Employment statusNo worker 25.0 10.7 42.7 91.0One worker 27.2 26.6 33.9 9.0
2
Two workers 36.1 46.4 19.6 0.0More than two workers 11.7 16.3 3.8 0.0Family typeSingle adult, no children 10.9 6.2 18.1 40.5Two adults, no children 42.8 49.9 43.2 14.6Single adult, children 4.6 1.1 6.1 21.7Two adults, children 32.4 35.3 21.3 17.3Large families 9.2 7.6 11.3 6.0Age of household headc
Young-age head 15.5 13.3 17.6 23.5Prime-age head 47.7 53.6 39.9 23.4Older-working-age head 20.4 22.0 23.0 11.6Retirement-age head 16.4 11.1 19.2 41.5Education leveld
Low education 33.1 24.4 34.7 63.9Middle education 36.9 38.0 38.9 29.1Higher education 30.0 37.6 26.4 7.0
UNITED STATESHead genderHead male 79.5 86.8 70.4 26.9Head female 20.5 13.2 29.7 73.1Employment statusNo worker 10.7 6.6 17.9 51.3One worker 32.3 29.0 52.6 40.3Two workers 42.4 47.6 25.6 7.0More than two workers 14.6 16.8 3.9 1.4Family typeSingle adult, no children 12.3 10.6 20.0 21.5Two adults, no children 29.8 34.3 21.6 8.4Single adult, children 8.1 3.8 12.3 44.5Two adults, children 35.5 37.7 29.8 14.4Large families 14.3 13.6 16.4 11.3Age of household headc
Young-age head 19.7 16.0 33.3 37.1Prime-age head 51.7 55.1 41.6 35.5Older-working-age head 19.1 20.6 15.8 12.1Retirement-age head 9.4 8.3 9.3 15.3Education leveld
Low education 18.9 12.2 24.9 57.4Middle education 36.9 35.6 43.2 31.6Higher education 44.2 52.2 31.8 11.1
Note: Includes corrected values which differ from those found in Eonomic Outlook 64, Table VI.3. For definitions see Oxley, Antolin and Dang (op. cit.).Characteristics refer to the household head.
a) Characteristics defined at the beginning of the period.b) Individuals who are poor only one year over the period, excluding spells occurring in the first and last year of the six-year period.c) Young, prime-age, older-working-age, and retirement-age refer, respectively, to households with a head below 30, between 30 and below 50,between 50 and 65, and above 65 years old.d) Low education is less than high-school; middle is completed high-school and higher is more than high-school education.Source: OECD Secretariat.
3
Table 4. Frequency of “events” associated with poverty transitionsPer cent share of total transitions
Entries ExitsTotal population Household with Total population Household with
working-age heada working-age heada
CanadaTransitions by typeb:Employment/earnings-related 26.1 28.1 38.4 44.4Family-structure-related 19.0 20.4 16.1 28.1Other factorsc 37.9 33.7 28.2 19.7
GermanyTransitions by typeb:Employment/earnings-related 47.1 51.5 50.1 55.0Family structure-related 24.7 25.2 8.9 9.5Other factorsc 23.4 17.9 33.3 27.1
United KingdomTransitions by typeb:Employment/earnings-related 27.5 34.2 41.2 51.6Family structure-related 26.4 28.5 9.1 10.2Other factorsc 35.4 23.4 40.5 26.6
United StatesTransitions by typeb:Employment/earnings-related 57.4 61.5 62.1 64.9Family structure-related 19.0 20.1 12.8 13.9Other factorsc 13.8 8.2 11.2 7.5
Note: See Oxley, Antolin and Dang (forthcoming) for description. Covers most recent six-year period of available data.a) Refers to individuals in households with a working-age head.b) The per cent shares do not sum to 100 per cent. The remainder includes cases which could not be classified.c) Households which were either employed or unemployed in both periods and where income changes were not dominated by movements in earnings.
Largely transfer-related.Source: OECD Secretariat.
4
Table 5. Changes in probability of entry and exit when an “event” occursChanges in probability points
Canadaa Germany United Kingdom United StatesPoverty entryb
Change in employment status onlyLoss of all workers 0.26 0.41 0.28 0.37Loss of some but not all workers 0.05 0.04 -0.01 0.08One worker, fall in hours – 0.15 0.12 0.22More than one worker, fall in hours – 0.00 -0.04 0.01
Change in family structure onlySeparations/divorce (spouse becomes head) 0.36 0.14 0.06 0.08Child becomes head – 0.09 0.09 0.16
Concomitant changes in employment and family statusSeparations/divorce (spouse becomes head)
Loss of all workers 0.72 0.82 0.49 0.62Loss of some but not all workers 0.47 0.31 0.05 0.25One worker, fall in hours – 0.58 0.28 0.45
Child becomes headLoss of all workers – 0.75 0.55 0.73Loss of some but not all workers – 0.23 0.07 0.38One worker, fall in hours – 0.48 0.34 0.59
Poverty exitb
Change in employment status onlyFrom zero to at least one worker 0.17 0.20 0.19 0.08Additional workers in working households 0.03 0.40 0.36 0.37One worker, increase in hours – 0.43 0.08 0.34More than one worker, increase in hours – – 0.30 0.47
Change in family structure onlyMarriage/partnership (head becomes spouse) 0.36 0.11 0.00 0.39Child becomes spouse – – 0.12 0.37
Concomitant changes in employment and family statusMarriage/partnership (head becomes spouse)
From zero to at least one worker 0.44 0.30 0.19 0.46Additional workers in working households 0.37 0.46 0.36 0.63One worker, increase in hours – 0.49 0.08 0.62
Child becomes headFrom zero to at least one worker – – 0.30 0.45Additional workers in working households – – 0.45 0.62One worker, increase in hours – – 0.20 0.61
a) Results for the Canadian data are less comparable to the other countries because information on labour-market and family status is more limited and someresults may be affected by breaks in the data.
b) Changes in probability points are defined relative to a baseline which is, essentially, the probability of transiting when there is no change either in familyor employment status.
Source: OECD Secretariat.
5
![[ESRC][K55]TankGame Project](https://img.pdfslide.net/doc/110x75/577cd6c01a28ab9e789d27bf/esrck55tankgame-project.jpg)
![ESRC Report [English]](https://img.pdfslide.net/doc/110x75/577d1fa91a28ab4e1e910c34/esrc-report-english.jpg)
